[DISCUSS] Enhance Support for Multicast Communication Pattern

[Yun Gao]

Hi everyone,
      In some scenarios we met a requirement that some operators want to send records to theirs downstream operators with an multicast communication pattern. In detail, for some records, the operators want to send them according to the partitioner (for example, Rebalance), and for some other records, the operators want to send them to all the connected operators and tasks. Such a communication pattern could be viewed as a kind of multicast: it does not broadcast every record, but some record will indeed be sent to multiple downstream operators.

However, we found that this kind of communication pattern seems could not be implemented rightly if the operators have multiple consumers with different parallelism, using the customized partitioner. To solve the above problem, we propose to enhance the support for such kind of irregular communication pattern. We think there may be two options:

     1. Support a kind of customized operator events, which share much similarity with Watermark, and these events can be broadcasted to the downstream operators separately.
     2. Let the channel selector supports multicast, and also add the separate RecordWriter implementation to avoid impacting the performance of the channel selector that does not need multicast.

The problem and options are detailed in https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing

We are also wondering if there are other methods to implement this requirement with or without changing Runtime. Very thanks for any feedbacks !


Best,
Yun

[Zhu Zhu]

Thanks Yun for starting this discussion.
I think the multicasting can be very helpful in certain cases.

I have received requirements from users that they want to do broadcast
join, while the data set to broadcast is too large to fit in one task.
Thus the requirement turned out to be to support cartesian product of 2
data set(one of which can be infinite stream).
For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
C.
The idea to is have 4 C subtasks to deal with different combinations of A/B
partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
This requires one record to be sent to multiple downstream subtasks, but
not to all subtasks.

With current interface this is not supported, as one record can only be
sent to one subtask, or to all subtasks of a JobVertex.
And the user had to split the broadcast data set manually to several
different JobVertices, which is hard to maintain and extend.

Thanks,
Zhu Zhu

Yun Gao <[hidden email]> 于2019年8月22日周四 下午8:42写道：

> Hi everyone,
>       In some scenarios we met a requirement that some operators want to
> send records to theirs downstream operators with an multicast communication
> pattern. In detail, for some records, the operators want to send them
> according to the partitioner (for example, Rebalance), and for some other
> records, the operators want to send them to all the connected operators and
> tasks. Such a communication pattern could be viewed as a kind of multicast:
> it does not broadcast every record, but some record will indeed be sent to
> multiple downstream operators.
>
> However, we found that this kind of communication pattern seems could not
> be implemented rightly if the operators have multiple consumers with
> different parallelism, using the customized partitioner. To solve the above
> problem, we propose to enhance the support for such kind of irregular
> communication pattern. We think there may be two options:
>
>      1. Support a kind of customized operator events, which share much
> similarity with Watermark, and these events can be broadcasted to the
> downstream operators separately.
>      2. Let the channel selector supports multicast, and also add the
> separate RecordWriter implementation to avoid impacting the performance of
> the channel selector that does not need multicast.
>
> The problem and options are detailed in
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>
> We are also wondering if there are other methods to implement this
> requirement with or without changing Runtime. Very thanks for any feedbacks
> !
>
>
> Best,
> Yun
>
>

[Piotr Nowojski-3]

Hi,

Yun:

Thanks for proposing the idea. I have checked the document and left couple of questions there, but it might be better to answer them here.

What is the exact motivation and what problems do you want to solve? We have dropped multicast support from the network stack [1] for two reasons:
1. Performance
2. Code simplicity

The proposal to re introduce `int[] ChannelSelector#selectChannels()` would revert those changes. At that time we were thinking about a way how to keep the multicast support on the network level, while keeping the performance and simplicity for non multicast cases and there are ways to achieve that. However they would add extra complexity to Flink, which it would be better to avoid.

On the other hand, supporting dual pattern: standard partitioning or broadcasting is easy to do, as LatencyMarkers are doing exactly that. It would be just a matter of exposing this to the user in some way. So before we go any further, can you describe your use cases/motivation? Isn’t mix of standard partitioning and broadcasting enough? Do we need multicasting?

Zhu:

Could you rephrase your example? I didn’t quite understand it.

Piotrek

[1] https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662>

> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email]> wrote:
>
> Thanks Yun for starting this discussion.
> I think the multicasting can be very helpful in certain cases.
>
> I have received requirements from users that they want to do broadcast
> join, while the data set to broadcast is too large to fit in one task.
> Thus the requirement turned out to be to support cartesian product of 2
> data set(one of which can be infinite stream).
> For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
> C.
> The idea to is have 4 C subtasks to deal with different combinations of A/B
> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> This requires one record to be sent to multiple downstream subtasks, but
> not to all subtasks.
>
> With current interface this is not supported, as one record can only be
> sent to one subtask, or to all subtasks of a JobVertex.
> And the user had to split the broadcast data set manually to several
> different JobVertices, which is hard to maintain and extend.
>
> Thanks,
> Zhu Zhu
>
> Yun Gao <[hidden email]> 于2019年8月22日周四 下午8:42写道：
>
>> Hi everyone,
>>      In some scenarios we met a requirement that some operators want to
>> send records to theirs downstream operators with an multicast communication
>> pattern. In detail, for some records, the operators want to send them
>> according to the partitioner (for example, Rebalance), and for some other
>> records, the operators want to send them to all the connected operators and
>> tasks. Such a communication pattern could be viewed as a kind of multicast:
>> it does not broadcast every record, but some record will indeed be sent to
>> multiple downstream operators.
>>
>> However, we found that this kind of communication pattern seems could not
>> be implemented rightly if the operators have multiple consumers with
>> different parallelism, using the customized partitioner. To solve the above
>> problem, we propose to enhance the support for such kind of irregular
>> communication pattern. We think there may be two options:
>>
>>     1. Support a kind of customized operator events, which share much
>> similarity with Watermark, and these events can be broadcasted to the
>> downstream operators separately.
>>     2. Let the channel selector supports multicast, and also add the
>> separate RecordWriter implementation to avoid impacting the performance of
>> the channel selector that does not need multicast.
>>
>> The problem and options are detailed in
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>>
>> We are also wondering if there are other methods to implement this
>> requirement with or without changing Runtime. Very thanks for any feedbacks
>> !
>>
>>
>> Best,
>> Yun
>>
>>

[Zhu Zhu]

Hi Piotr,

The case is about a broadcast join:
A--\
     +--(join)--> C
B--/

Usually we can broadcast A(the result that JobVertex A produces) to all
subtasks of C.
But in this case the size of A is too large to fit in one subtask of C.
Thus we have to partition A to (A_0, A_1, A_2, ..., A_m-1).
The throughput of B is too large to deal in one subtask as well. And we
partition B into (B_0, B_1, B_2, ..., B_n-1).

Now if we want to join A and B, the basic idea is to set parallelism of C
to be m*n, and subtask C_kn+l should deal with the join work of (A_k, B_l).
To achieve this,
each record in partition A_k should to sent to *n* downstream subtasks:
{C_kn, C_kn+1, C_kn+2, ..., C_kn+n-1}
each record in partition B_l should to sent to *m* downstream
subtasks:  {C_l, C_n+l, C_2n+l, ..., C_(m-1)n+l}

This is different from current single-cast or broad-cast way.
That's why I think multi-cast can help with this case.

Thanks,
Zhu Zhu

Piotr Nowojski <[hidden email]> 于2019年8月23日周五 下午3:20写道：

> Hi,
>
> Yun:
>
> Thanks for proposing the idea. I have checked the document and left couple
> of questions there, but it might be better to answer them here.
>
> What is the exact motivation and what problems do you want to solve? We
> have dropped multicast support from the network stack [1] for two reasons:
> 1. Performance
> 2. Code simplicity
>
> The proposal to re introduce `int[] ChannelSelector#selectChannels()`
> would revert those changes. At that time we were thinking about a way how
> to keep the multicast support on the network level, while keeping the
> performance and simplicity for non multicast cases and there are ways to
> achieve that. However they would add extra complexity to Flink, which it
> would be better to avoid.
>
> On the other hand, supporting dual pattern: standard partitioning or
> broadcasting is easy to do, as LatencyMarkers are doing exactly that. It
> would be just a matter of exposing this to the user in some way. So before
> we go any further, can you describe your use cases/motivation? Isn’t mix of
> standard partitioning and broadcasting enough? Do we need multicasting?
>
> Zhu:
>
> Could you rephrase your example? I didn’t quite understand it.
>
> Piotrek
>
> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> https://issues.apache.org/jira/browse/FLINK-10662>
>
> > On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email]> wrote:
> >
> > Thanks Yun for starting this discussion.
> > I think the multicasting can be very helpful in certain cases.
> >
> > I have received requirements from users that they want to do broadcast
> > join, while the data set to broadcast is too large to fit in one task.
> > Thus the requirement turned out to be to support cartesian product of 2
> > data set(one of which can be infinite stream).
> > For example, A(parallelism=2) broadcast join B(parallelism=2) in
> JobVertex
> > C.
> > The idea to is have 4 C subtasks to deal with different combinations of
> A/B
> > partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> > This requires one record to be sent to multiple downstream subtasks, but
> > not to all subtasks.
> >
> > With current interface this is not supported, as one record can only be
> > sent to one subtask, or to all subtasks of a JobVertex.
> > And the user had to split the broadcast data set manually to several
> > different JobVertices, which is hard to maintain and extend.
> >
> > Thanks,
> > Zhu Zhu
> >
> > Yun Gao <[hidden email]> 于2019年8月22日周四 下午8:42写道：
> >
> >> Hi everyone,
> >>      In some scenarios we met a requirement that some operators want to
> >> send records to theirs downstream operators with an multicast
> communication
> >> pattern. In detail, for some records, the operators want to send them
> >> according to the partitioner (for example, Rebalance), and for some
> other
> >> records, the operators want to send them to all the connected operators
> and
> >> tasks. Such a communication pattern could be viewed as a kind of
> multicast:
> >> it does not broadcast every record, but some record will indeed be sent
> to
> >> multiple downstream operators.
> >>
> >> However, we found that this kind of communication pattern seems could
> not
> >> be implemented rightly if the operators have multiple consumers with
> >> different parallelism, using the customized partitioner. To solve the
> above
> >> problem, we propose to enhance the support for such kind of irregular
> >> communication pattern. We think there may be two options:
> >>
> >>     1. Support a kind of customized operator events, which share much
> >> similarity with Watermark, and these events can be broadcasted to the
> >> downstream operators separately.
> >>     2. Let the channel selector supports multicast, and also add the
> >> separate RecordWriter implementation to avoid impacting the performance
> of
> >> the channel selector that does not need multicast.
> >>
> >> The problem and options are detailed in
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >>
> >> We are also wondering if there are other methods to implement this
> >> requirement with or without changing Runtime. Very thanks for any
> feedbacks
> >> !
> >>
> >>
> >> Best,
> >> Yun
> >>
> >>
>
>

[Yun Gao]

   Hi Piotr,

        Thanks a lot for the suggestions!

        The core motivation of this discussion is to implement a new iteration library on the DataStream, and it requires to insert special records in the stream to notify the progress of the iteration. The mechanism of such records is very similar to the current Watermark, and we meet the problem of sending normal records according to the partition (Rebalance, etc..) and also be able to broadcast the inserted progress records to all the connected records. I have read the notes in the google doc and I totally agree with that exploring the broadcast interface in RecordWriter in some way is able to solve this issue.

       Regarding to `int[] ChannelSelector#selectChannels()`, I'm wondering if we introduce a new MulticastRecordWriter and left the current RecordWriter untouched, could we avoid the performance degradation ? Since with such a modification the normal RecordWriter does not need to iterate the return array by ChannelSelector, and the only difference will be returning an array instead of an integer, and accessing the first element of the returned array instead of reading the integer directly.

Best,
Yun


------------------------------------------------------------------
From:Piotr Nowojski <[hidden email]>
Send Time:2019 Aug. 23 (Fri.) 15:20
To:dev <[hidden email]>
Cc:Yun Gao <[hidden email]>
Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern

Hi,

Yun:

Thanks for proposing the idea. I have checked the document and left couple of questions there, but it might be better to answer them here.

What is the exact motivation and what problems do you want to solve? We have dropped multicast support from the network stack [1] for two reasons:
1. Performance
2. Code simplicity

The proposal to re introduce `int[] ChannelSelector#selectChannels()` would revert those changes. At that time we were thinking about a way how to keep the multicast support on the network level, while keeping the performance and simplicity for non multicast cases and there are ways to achieve that. However they would add extra complexity to Flink, which it would be better to avoid.

On the other hand, supporting dual pattern: standard partitioning or broadcasting is easy to do, as LatencyMarkers are doing exactly that. It would be just a matter of exposing this to the user in some way. So before we go any further, can you describe your use cases/motivation? Isn’t mix of standard partitioning and broadcasting enough? Do we need multicasting?

Zhu:

Could you rephrase your example? I didn’t quite understand it.

Piotrek

[1] https://issues.apache.org/jira/browse/FLINK-10662


On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email]> wrote:
Thanks Yun for starting this discussion.
I think the multicasting can be very helpful in certain cases.

I have received requirements from users that they want to do broadcast
join, while the data set to broadcast is too large to fit in one task.
Thus the requirement turned out to be to support cartesian product of 2
data set(one of which can be infinite stream).
For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
C.
The idea to is have 4 C subtasks to deal with different combinations of A/B
partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
This requires one record to be sent to multiple downstream subtasks, but
not to all subtasks.

With current interface this is not supported, as one record can only be
sent to one subtask, or to all subtasks of a JobVertex.
And the user had to split the broadcast data set manually to several
different JobVertices, which is hard to maintain and extend.

Thanks,
Zhu Zhu

Yun Gao <[hidden email]> 于2019年8月22日周四 下午8:42写道：

Hi everyone,
      In some scenarios we met a requirement that some operators want to
send records to theirs downstream operators with an multicast communication
pattern. In detail, for some records, the operators want to send them
according to the partitioner (for example, Rebalance), and for some other
records, the operators want to send them to all the connected operators and
tasks. Such a communication pattern could be viewed as a kind of multicast:
it does not broadcast every record, but some record will indeed be sent to
multiple downstream operators.

However, we found that this kind of communication pattern seems could not
be implemented rightly if the operators have multiple consumers with
different parallelism, using the customized partitioner. To solve the above
problem, we propose to enhance the support for such kind of irregular
communication pattern. We think there may be two options:

     1. Support a kind of customized operator events, which share much
similarity with Watermark, and these events can be broadcasted to the
downstream operators separately.
     2. Let the channel selector supports multicast, and also add the
separate RecordWriter implementation to avoid impacting the performance of
the channel selector that does not need multicast.

The problem and options are detailed in
https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing

We are also wondering if there are other methods to implement this
requirement with or without changing Runtime. Very thanks for any feedbacks
!


Best,
Yun

[Piotr Nowojski-3]

Hi Yun and Zhu Zhu,

Thanks for the more detailed example Zhu Zhu.

As far as I understand for the iterations example we do not need multicasting. Regarding the Join example, I don’t fully understand it. The example that Zhu Zhu presented has a drawback of sending both tables to multiple nodes. What’s the benefit of using broadcast join over a hash join in such case? As far as I know, the biggest benefit of using broadcast join instead of hash join is that we can avoid sending the larger table over the network, because we can perform the join locally. In this example we are sending both of the tables to multiple nodes, which should defeat the purpose.

Is it about implementing cross join or near cross joins in a distributed fashion?

> if we introduce a new MulticastRecordWriter

That’s one of the solutions. It might have a drawback of 3 class virtualisation problem (We have RecordWriter and BroadcastRecordWriter already). With up to two implementations, JVM is able to devirtualise the calls.

Previously I was also thinking about just providing two different ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector` with plain `int` and based on that, RecordWriter could perform some magic (worst case scenario `instaceof` checks).

Another solution might be to change `ChannelSelector` interface into an iterator.

But let's discuss the details after we agree on implementing this.

Piotrek

> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email]> wrote:
>
>    Hi Piotr,
>
>         Thanks a lot for the suggestions!
>
>         The core motivation of this discussion is to implement a new iteration library on the DataStream, and it requires to insert special records in the stream to notify the progress of the iteration. The mechanism of such records is very similar to the current Watermark, and we meet the problem of sending normal records according to the partition (Rebalance, etc..) and also be able to broadcast the inserted progress records to all the connected records. I have read the notes in the google doc and I totally agree with that exploring the broadcast interface in RecordWriter in some way is able to solve this issue.
>
>        Regarding to `int[] ChannelSelector#selectChannels()`, I'm wondering if we introduce a new MulticastRecordWriter and left the current RecordWriter untouched, could we avoid the performance degradation ? Since with such a modification the normal RecordWriter does not need to iterate the return array by ChannelSelector, and the only difference will be returning an array instead of an integer, and accessing the first element of the returned array instead of reading the integer directly.
>
> Best,
> Yun
>
> ------------------------------------------------------------------
> From:Piotr Nowojski <[hidden email]>
> Send Time:2019 Aug. 23 (Fri.) 15:20
> To:dev <[hidden email]>
> Cc:Yun Gao <[hidden email]>
> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
>
> Hi,
>
> Yun:
>
> Thanks for proposing the idea. I have checked the document and left couple of questions there, but it might be better to answer them here.
>
> What is the exact motivation and what problems do you want to solve? We have dropped multicast support from the network stack [1] for two reasons:
> 1. Performance
> 2. Code simplicity
>
> The proposal to re introduce `int[] ChannelSelector#selectChannels()` would revert those changes. At that time we were thinking about a way how to keep the multicast support on the network level, while keeping the performance and simplicity for non multicast cases and there are ways to achieve that. However they would add extra complexity to Flink, which it would be better to avoid.
>
> On the other hand, supporting dual pattern: standard partitioning or broadcasting is easy to do, as LatencyMarkers are doing exactly that. It would be just a matter of exposing this to the user in some way. So before we go any further, can you describe your use cases/motivation? Isn’t mix of standard partitioning and broadcasting enough? Do we need multicasting?
>
> Zhu:
>
> Could you rephrase your example? I didn’t quite understand it.
>
> Piotrek
>
> [1] https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662>
>
> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:[hidden email]>> wrote:
>
> Thanks Yun for starting this discussion.
> I think the multicasting can be very helpful in certain cases.
>
> I have received requirements from users that they want to do broadcast
> join, while the data set to broadcast is too large to fit in one task.
> Thus the requirement turned out to be to support cartesian product of 2
> data set(one of which can be infinite stream).
> For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
> C.
> The idea to is have 4 C subtasks to deal with different combinations of A/B
> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> This requires one record to be sent to multiple downstream subtasks, but
> not to all subtasks.
>
> With current interface this is not supported, as one record can only be
> sent to one subtask, or to all subtasks of a JobVertex.
> And the user had to split the broadcast data set manually to several
> different JobVertices, which is hard to maintain and extend.
>
> Thanks,
> Zhu Zhu
>
> Yun Gao <[hidden email] <mailto:[hidden email]>> 于2019年8月22日周四 下午8:42写道：
>
> Hi everyone,
>      In some scenarios we met a requirement that some operators want to
> send records to theirs downstream operators with an multicast communication
> pattern. In detail, for some records, the operators want to send them
> according to the partitioner (for example, Rebalance), and for some other
> records, the operators want to send them to all the connected operators and
> tasks. Such a communication pattern could be viewed as a kind of multicast:
> it does not broadcast every record, but some record will indeed be sent to
> multiple downstream operators.
>
> However, we found that this kind of communication pattern seems could not
> be implemented rightly if the operators have multiple consumers with
> different parallelism, using the customized partitioner. To solve the above
> problem, we propose to enhance the support for such kind of irregular
> communication pattern. We think there may be two options:
>
>     1. Support a kind of customized operator events, which share much
> similarity with Watermark, and these events can be broadcasted to the
> downstream operators separately.
>     2. Let the channel selector supports multicast, and also add the
> separate RecordWriter implementation to avoid impacting the performance of
> the channel selector that does not need multicast.
>
> The problem and options are detailed in
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing>
>
> We are also wondering if there are other methods to implement this
> requirement with or without changing Runtime. Very thanks for any feedbacks
> !
>
>
> Best,
> Yun
>
>
>
>

[Zhu Zhu]

Hi Piotr,

Yes you are right it's a distributed cross join requirement.
Broadcast join can help with cross join cases. But users cannot use it if
the data set to join is too large to fit into one subtask.

Sorry for left some details behind.

Thanks,
Zhu Zhu

Piotr Nowojski <[hidden email]> 于2019年8月23日周五 下午4:57写道：

> Hi Yun and Zhu Zhu,
>
> Thanks for the more detailed example Zhu Zhu.
>
> As far as I understand for the iterations example we do not need
> multicasting. Regarding the Join example, I don’t fully understand it. The
> example that Zhu Zhu presented has a drawback of sending both tables to
> multiple nodes. What’s the benefit of using broadcast join over a hash join
> in such case? As far as I know, the biggest benefit of using broadcast join
> instead of hash join is that we can avoid sending the larger table over the
> network, because we can perform the join locally. In this example we are
> sending both of the tables to multiple nodes, which should defeat the
> purpose.
>
> Is it about implementing cross join or near cross joins in a distributed
> fashion?
>
> > if we introduce a new MulticastRecordWriter
>
> That’s one of the solutions. It might have a drawback of 3 class
> virtualisation problem (We have RecordWriter and BroadcastRecordWriter
> already). With up to two implementations, JVM is able to devirtualise the
> calls.
>
> Previously I was also thinking about just providing two different
> ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector`
> with plain `int` and based on that, RecordWriter could perform some magic
> (worst case scenario `instaceof` checks).
>
> Another solution might be to change `ChannelSelector` interface into an
> iterator.
>
> But let's discuss the details after we agree on implementing this.
>
> Piotrek
>
> > On 23 Aug 2019, at 10:20, Yun Gao <[hidden email]> wrote:
> >
> >    Hi Piotr,
> >
> >         Thanks a lot for the suggestions!
> >
> >         The core motivation of this discussion is to implement a new
> iteration library on the DataStream, and it requires to insert special
> records in the stream to notify the progress of the iteration. The
> mechanism of such records is very similar to the current Watermark, and we
> meet the problem of sending normal records according to the partition
> (Rebalance, etc..) and also be able to broadcast the inserted progress
> records to all the connected records. I have read the notes in the google
> doc and I totally agree with that exploring the broadcast interface in
> RecordWriter in some way is able to solve this issue.
> >
> >        Regarding to `int[] ChannelSelector#selectChannels()`, I'm
> wondering if we introduce a new MulticastRecordWriter and left the current
> RecordWriter untouched, could we avoid the performance degradation ? Since
> with such a modification the normal RecordWriter does not need to iterate
> the return array by ChannelSelector, and the only difference will be
> returning an array instead of an integer, and accessing the first element
> of the returned array instead of reading the integer directly.
> >
> > Best,
> > Yun
> >
> > ------------------------------------------------------------------
> > From:Piotr Nowojski <[hidden email]>
> > Send Time:2019 Aug. 23 (Fri.) 15:20
> > To:dev <[hidden email]>
> > Cc:Yun Gao <[hidden email]>
> > Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
> >
> > Hi,
> >
> > Yun:
> >
> > Thanks for proposing the idea. I have checked the document and left
> couple of questions there, but it might be better to answer them here.
> >
> > What is the exact motivation and what problems do you want to solve? We
> have dropped multicast support from the network stack [1] for two reasons:
> > 1. Performance
> > 2. Code simplicity
> >
> > The proposal to re introduce `int[] ChannelSelector#selectChannels()`
> would revert those changes. At that time we were thinking about a way how
> to keep the multicast support on the network level, while keeping the
> performance and simplicity for non multicast cases and there are ways to
> achieve that. However they would add extra complexity to Flink, which it
> would be better to avoid.
> >
> > On the other hand, supporting dual pattern: standard partitioning or
> broadcasting is easy to do, as LatencyMarkers are doing exactly that. It
> would be just a matter of exposing this to the user in some way. So before
> we go any further, can you describe your use cases/motivation? Isn’t mix of
> standard partitioning and broadcasting enough? Do we need multicasting?
> >
> > Zhu:
> >
> > Could you rephrase your example? I didn’t quite understand it.
> >
> > Piotrek
> >
> > [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> https://issues.apache.org/jira/browse/FLINK-10662>
> >
> > On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
> [hidden email]>> wrote:
> >
> > Thanks Yun for starting this discussion.
> > I think the multicasting can be very helpful in certain cases.
> >
> > I have received requirements from users that they want to do broadcast
> > join, while the data set to broadcast is too large to fit in one task.
> > Thus the requirement turned out to be to support cartesian product of 2
> > data set(one of which can be infinite stream).
> > For example, A(parallelism=2) broadcast join B(parallelism=2) in
> JobVertex
> > C.
> > The idea to is have 4 C subtasks to deal with different combinations of
> A/B
> > partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> > This requires one record to be sent to multiple downstream subtasks, but
> > not to all subtasks.
> >
> > With current interface this is not supported, as one record can only be
> > sent to one subtask, or to all subtasks of a JobVertex.
> > And the user had to split the broadcast data set manually to several
> > different JobVertices, which is hard to maintain and extend.
> >
> > Thanks,
> > Zhu Zhu
> >
> > Yun Gao <[hidden email] <mailto:
> [hidden email]>> 于2019年8月22日周四 下午8:42写道：
> >
> > Hi everyone,
> >      In some scenarios we met a requirement that some operators want to
> > send records to theirs downstream operators with an multicast
> communication
> > pattern. In detail, for some records, the operators want to send them
> > according to the partitioner (for example, Rebalance), and for some other
> > records, the operators want to send them to all the connected operators
> and
> > tasks. Such a communication pattern could be viewed as a kind of
> multicast:
> > it does not broadcast every record, but some record will indeed be sent
> to
> > multiple downstream operators.
> >
> > However, we found that this kind of communication pattern seems could not
> > be implemented rightly if the operators have multiple consumers with
> > different parallelism, using the customized partitioner. To solve the
> above
> > problem, we propose to enhance the support for such kind of irregular
> > communication pattern. We think there may be two options:
> >
> >     1. Support a kind of customized operator events, which share much
> > similarity with Watermark, and these events can be broadcasted to the
> > downstream operators separately.
> >     2. Let the channel selector supports multicast, and also add the
> > separate RecordWriter implementation to avoid impacting the performance
> of
> > the channel selector that does not need multicast.
> >
> > The problem and options are detailed in
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> <
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >
> >
> > We are also wondering if there are other methods to implement this
> > requirement with or without changing Runtime. Very thanks for any
> feedbacks
> > !
> >
> >
> > Best,
> > Yun
> >
> >
> >
> >
>
>

[Yun Gao]

     Hi Piotr,

        Thanks a lot for sharing the thoughts!

        For the iteration, agree with that multicasting is not necessary. Exploring the broadcast interface to Output of the operators in some way should also solve this issue, and I think it should be even more convenient to have the broadcast method for the iteration.

        Also thanks Zhu Zhu for the cross join case!
  Best,
   Yun



------------------------------------------------------------------
From:Zhu Zhu <[hidden email]>
Send Time:2019 Aug. 23 (Fri.) 17:25
To:dev <[hidden email]>
Cc:Yun Gao <[hidden email]>
Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern

Hi Piotr,

Yes you are right it's a distributed cross join requirement.
Broadcast join can help with cross join cases. But users cannot use it if the data set to join is too large to fit into one subtask.

Sorry for left some details behind.

Thanks,
Zhu Zhu
Piotr Nowojski <[hidden email]> 于2019年8月23日周五 下午4:57写道：
Hi Yun and Zhu Zhu,

 Thanks for the more detailed example Zhu Zhu.

 As far as I understand for the iterations example we do not need multicasting. Regarding the Join example, I don’t fully understand it. The example that Zhu Zhu presented has a drawback of sending both tables to multiple nodes. What’s the benefit of using broadcast join over a hash join in such case? As far as I know, the biggest benefit of using broadcast join instead of hash join is that we can avoid sending the larger table over the network, because we can perform the join locally. In this example we are sending both of the tables to multiple nodes, which should defeat the purpose.

 Is it about implementing cross join or near cross joins in a distributed fashion?

 > if we introduce a new MulticastRecordWriter

 That’s one of the solutions. It might have a drawback of 3 class virtualisation problem (We have RecordWriter and BroadcastRecordWriter already). With up to two implementations, JVM is able to devirtualise the calls.

 Previously I was also thinking about just providing two different ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector` with plain `int` and based on that, RecordWriter could perform some magic (worst case scenario `instaceof` checks).

 Another solution might be to change `ChannelSelector` interface into an iterator.

 But let's discuss the details after we agree on implementing this.

 Piotrek

 > On 23 Aug 2019, at 10:20, Yun Gao <[hidden email]> wrote:
 >
 >    Hi Piotr,
 >
 >         Thanks a lot for the suggestions!
 >
 >         The core motivation of this discussion is to implement a new iteration library on the DataStream, and it requires to insert special records in the stream to notify the progress of the iteration. The mechanism of such records is very similar to the current Watermark, and we meet the problem of sending normal records according to the partition (Rebalance, etc..) and also be able to broadcast the inserted progress records to all the connected records. I have read the notes in the google doc and I totally agree with that exploring the broadcast interface in RecordWriter in some way is able to solve this issue.
 >
 >        Regarding to `int[] ChannelSelector#selectChannels()`, I'm wondering if we introduce a new MulticastRecordWriter and left the current RecordWriter untouched, could we avoid the performance degradation ? Since with such a modification the normal RecordWriter does not need to iterate the return array by ChannelSelector, and the only difference will be returning an array instead of an integer, and accessing the first element of the returned array instead of reading the integer directly.
 >
 > Best,
 > Yun
 >
 > ------------------------------------------------------------------
 > From:Piotr Nowojski <[hidden email]>
 > Send Time:2019 Aug. 23 (Fri.) 15:20
 > To:dev <[hidden email]>
 > Cc:Yun Gao <[hidden email]>
 > Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
 >
 > Hi,
 >
 > Yun:
 >
 > Thanks for proposing the idea. I have checked the document and left couple of questions there, but it might be better to answer them here.
 >
 > What is the exact motivation and what problems do you want to solve? We have dropped multicast support from the network stack [1] for two reasons:
 > 1. Performance
 > 2. Code simplicity
 >
 > The proposal to re introduce `int[] ChannelSelector#selectChannels()` would revert those changes. At that time we were thinking about a way how to keep the multicast support on the network level, while keeping the performance and simplicity for non multicast cases and there are ways to achieve that. However they would add extra complexity to Flink, which it would be better to avoid.
 >
 > On the other hand, supporting dual pattern: standard partitioning or broadcasting is easy to do, as LatencyMarkers are doing exactly that. It would be just a matter of exposing this to the user in some way. So before we go any further, can you describe your use cases/motivation? Isn’t mix of standard partitioning and broadcasting enough? Do we need multicasting?
 >
 > Zhu:
 >
 > Could you rephrase your example? I didn’t quite understand it.
 >
 > Piotrek
 >
 > [1] https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662>
 >
 > On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:[hidden email]>> wrote:
 >
 > Thanks Yun for starting this discussion.
 > I think the multicasting can be very helpful in certain cases.
 >
 > I have received requirements from users that they want to do broadcast
 > join, while the data set to broadcast is too large to fit in one task.
 > Thus the requirement turned out to be to support cartesian product of 2
 > data set(one of which can be infinite stream).
 > For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
 > C.
 > The idea to is have 4 C subtasks to deal with different combinations of A/B
 > partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
 > This requires one record to be sent to multiple downstream subtasks, but
 > not to all subtasks.
 >
 > With current interface this is not supported, as one record can only be
 > sent to one subtask, or to all subtasks of a JobVertex.
 > And the user had to split the broadcast data set manually to several
 > different JobVertices, which is hard to maintain and extend.
 >
 > Thanks,
 > Zhu Zhu
 >
 > Yun Gao <[hidden email] <mailto:[hidden email]>> 于2019年8月22日周四 下午8:42写道：
 >
 > Hi everyone,
 >      In some scenarios we met a requirement that some operators want to
 > send records to theirs downstream operators with an multicast communication
 > pattern. In detail, for some records, the operators want to send them
 > according to the partitioner (for example, Rebalance), and for some other
 > records, the operators want to send them to all the connected operators and
 > tasks. Such a communication pattern could be viewed as a kind of multicast:
 > it does not broadcast every record, but some record will indeed be sent to
 > multiple downstream operators.
 >
 > However, we found that this kind of communication pattern seems could not
 > be implemented rightly if the operators have multiple consumers with
 > different parallelism, using the customized partitioner. To solve the above
 > problem, we propose to enhance the support for such kind of irregular
 > communication pattern. We think there may be two options:
 >
 >     1. Support a kind of customized operator events, which share much
 > similarity with Watermark, and these events can be broadcasted to the
 > downstream operators separately.
 >     2. Let the channel selector supports multicast, and also add the
 > separate RecordWriter implementation to avoid impacting the performance of
 > the channel selector that does not need multicast.
 >
 > The problem and options are detailed in
 > https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing>
 >
 > We are also wondering if there are other methods to implement this
 > requirement with or without changing Runtime. Very thanks for any feedbacks
 > !
 >
 >
 > Best,
 > Yun
 >
 >
 >
 >

[Piotr Nowojski-3]

Hi,

Thanks for the answers :) Ok I understand the full picture now. +1 from my side on solving this issue somehow. But before we start discussing how to solve it one last control question:

I guess this multicast is intended to be used in blink planner, right? Assuming that we implement the multicast support now, when would it be used by the blink? I would like to avoid a scenario, where we implement an unused feature and we keep maintaining it for a long period of time.

Piotrek

PS, try to include motivating examples, including concrete ones in the proposals/design docs, for example in the very first paragraph. Especially if it’s a commonly known feature like cross join :)

> On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]> wrote:
>
>     Hi Piotr,
>
>        Thanks a lot for sharing the thoughts!
>
>        For the iteration, agree with that multicasting is not necessary. Exploring the broadcast interface to Output of the operators in some way should also solve this issue, and I think it should be even more convenient to have the broadcast method for the iteration.
>
>        Also thanks Zhu Zhu for the cross join case!
>  Best,
>   Yun
>
>
>
> ------------------------------------------------------------------
> From:Zhu Zhu <[hidden email]>
> Send Time:2019 Aug. 23 (Fri.) 17:25
> To:dev <[hidden email]>
> Cc:Yun Gao <[hidden email]>
> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
>
> Hi Piotr,
>
> Yes you are right it's a distributed cross join requirement.
> Broadcast join can help with cross join cases. But users cannot use it if the data set to join is too large to fit into one subtask.
>
> Sorry for left some details behind.
>
> Thanks,
> Zhu Zhu
> Piotr Nowojski <[hidden email]> 于2019年8月23日周五 下午4:57写道：
> Hi Yun and Zhu Zhu,
>
> Thanks for the more detailed example Zhu Zhu.
>
> As far as I understand for the iterations example we do not need multicasting. Regarding the Join example, I don’t fully understand it. The example that Zhu Zhu presented has a drawback of sending both tables to multiple nodes. What’s the benefit of using broadcast join over a hash join in such case? As far as I know, the biggest benefit of using broadcast join instead of hash join is that we can avoid sending the larger table over the network, because we can perform the join locally. In this example we are sending both of the tables to multiple nodes, which should defeat the purpose.
>
> Is it about implementing cross join or near cross joins in a distributed fashion?
>
>> if we introduce a new MulticastRecordWriter
>
> That’s one of the solutions. It might have a drawback of 3 class virtualisation problem (We have RecordWriter and BroadcastRecordWriter already). With up to two implementations, JVM is able to devirtualise the calls.
>
> Previously I was also thinking about just providing two different ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector` with plain `int` and based on that, RecordWriter could perform some magic (worst case scenario `instaceof` checks).
>
> Another solution might be to change `ChannelSelector` interface into an iterator.
>
> But let's discuss the details after we agree on implementing this.
>
> Piotrek
>
>> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email]> wrote:
>>
>>   Hi Piotr,
>>
>>        Thanks a lot for the suggestions!
>>
>>        The core motivation of this discussion is to implement a new iteration library on the DataStream, and it requires to insert special records in the stream to notify the progress of the iteration. The mechanism of such records is very similar to the current Watermark, and we meet the problem of sending normal records according to the partition (Rebalance, etc..) and also be able to broadcast the inserted progress records to all the connected records. I have read the notes in the google doc and I totally agree with that exploring the broadcast interface in RecordWriter in some way is able to solve this issue.
>>
>>       Regarding to `int[] ChannelSelector#selectChannels()`, I'm wondering if we introduce a new MulticastRecordWriter and left the current RecordWriter untouched, could we avoid the performance degradation ? Since with such a modification the normal RecordWriter does not need to iterate the return array by ChannelSelector, and the only difference will be returning an array instead of an integer, and accessing the first element of the returned array instead of reading the integer directly.
>>
>> Best,
>> Yun
>>
>> ------------------------------------------------------------------
>> From:Piotr Nowojski <[hidden email]>
>> Send Time:2019 Aug. 23 (Fri.) 15:20
>> To:dev <[hidden email]>
>> Cc:Yun Gao <[hidden email]>
>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
>>
>> Hi,
>>
>> Yun:
>>
>> Thanks for proposing the idea. I have checked the document and left couple of questions there, but it might be better to answer them here.
>>
>> What is the exact motivation and what problems do you want to solve? We have dropped multicast support from the network stack [1] for two reasons:
>> 1. Performance
>> 2. Code simplicity
>>
>> The proposal to re introduce `int[] ChannelSelector#selectChannels()` would revert those changes. At that time we were thinking about a way how to keep the multicast support on the network level, while keeping the performance and simplicity for non multicast cases and there are ways to achieve that. However they would add extra complexity to Flink, which it would be better to avoid.
>>
>> On the other hand, supporting dual pattern: standard partitioning or broadcasting is easy to do, as LatencyMarkers are doing exactly that. It would be just a matter of exposing this to the user in some way. So before we go any further, can you describe your use cases/motivation? Isn’t mix of standard partitioning and broadcasting enough? Do we need multicasting?
>>
>> Zhu:
>>
>> Could you rephrase your example? I didn’t quite understand it.
>>
>> Piotrek
>>
>> [1] https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662>
>>
>> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:[hidden email]>> wrote:
>>
>> Thanks Yun for starting this discussion.
>> I think the multicasting can be very helpful in certain cases.
>>
>> I have received requirements from users that they want to do broadcast
>> join, while the data set to broadcast is too large to fit in one task.
>> Thus the requirement turned out to be to support cartesian product of 2
>> data set(one of which can be infinite stream).
>> For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
>> C.
>> The idea to is have 4 C subtasks to deal with different combinations of A/B
>> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
>> This requires one record to be sent to multiple downstream subtasks, but
>> not to all subtasks.
>>
>> With current interface this is not supported, as one record can only be
>> sent to one subtask, or to all subtasks of a JobVertex.
>> And the user had to split the broadcast data set manually to several
>> different JobVertices, which is hard to maintain and extend.
>>
>> Thanks,
>> Zhu Zhu
>>
>> Yun Gao <[hidden email] <mailto:[hidden email]>> 于2019年8月22日周四 下午8:42写道：
>>
>> Hi everyone,
>>     In some scenarios we met a requirement that some operators want to
>> send records to theirs downstream operators with an multicast communication
>> pattern. In detail, for some records, the operators want to send them
>> according to the partitioner (for example, Rebalance), and for some other
>> records, the operators want to send them to all the connected operators and
>> tasks. Such a communication pattern could be viewed as a kind of multicast:
>> it does not broadcast every record, but some record will indeed be sent to
>> multiple downstream operators.
>>
>> However, we found that this kind of communication pattern seems could not
>> be implemented rightly if the operators have multiple consumers with
>> different parallelism, using the customized partitioner. To solve the above
>> problem, we propose to enhance the support for such kind of irregular
>> communication pattern. We think there may be two options:
>>
>>    1. Support a kind of customized operator events, which share much
>> similarity with Watermark, and these events can be broadcasted to the
>> downstream operators separately.
>>    2. Let the channel selector supports multicast, and also add the
>> separate RecordWriter implementation to avoid impacting the performance of
>> the channel selector that does not need multicast.
>>
>> The problem and options are detailed in
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing>
>>
>> We are also wondering if there are other methods to implement this
>> requirement with or without changing Runtime. Very thanks for any feedbacks
>> !
>>
>>
>> Best,
>> Yun
>>
>>
>>
>>
>
>

[Zhu Zhu]

Thanks Piotr,

Users asked for this feature sometimes ago when they migrating batch jobs
to Flink(Blink).
It's not very urgent as they have taken some workarounds to solve it.(like
partitioning data set to different job vertices)
So it's fine to not make it top priority.

Anyway, as a commonly known scenario, I think users can benefit from cross
join sooner or later.

Thanks,
Zhu Zhu

Piotr Nowojski <[hidden email]> 于2019年8月23日周五 下午6:19写道：

> Hi,
>
> Thanks for the answers :) Ok I understand the full picture now. +1 from my
> side on solving this issue somehow. But before we start discussing how to
> solve it one last control question:
>
> I guess this multicast is intended to be used in blink planner, right?
> Assuming that we implement the multicast support now, when would it be used
> by the blink? I would like to avoid a scenario, where we implement an
> unused feature and we keep maintaining it for a long period of time.
>
> Piotrek
>
> PS, try to include motivating examples, including concrete ones in the
> proposals/design docs, for example in the very first paragraph. Especially
> if it’s a commonly known feature like cross join :)
>
> > On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]> wrote:
> >
> >     Hi Piotr,
> >
> >        Thanks a lot for sharing the thoughts!
> >
> >        For the iteration, agree with that multicasting is not necessary.
> Exploring the broadcast interface to Output of the operators in some way
> should also solve this issue, and I think it should be even more convenient
> to have the broadcast method for the iteration.
> >
> >        Also thanks Zhu Zhu for the cross join case!
> >  Best,
> >   Yun
> >
> >
> >
> > ------------------------------------------------------------------
> > From:Zhu Zhu <[hidden email]>
> > Send Time:2019 Aug. 23 (Fri.) 17:25
> > To:dev <[hidden email]>
> > Cc:Yun Gao <[hidden email]>
> > Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
> >
> > Hi Piotr,
> >
> > Yes you are right it's a distributed cross join requirement.
> > Broadcast join can help with cross join cases. But users cannot use it
> if the data set to join is too large to fit into one subtask.
> >
> > Sorry for left some details behind.
> >
> > Thanks,
> > Zhu Zhu
> > Piotr Nowojski <[hidden email]> 于2019年8月23日周五 下午4:57写道：
> > Hi Yun and Zhu Zhu,
> >
> > Thanks for the more detailed example Zhu Zhu.
> >
> > As far as I understand for the iterations example we do not need
> multicasting. Regarding the Join example, I don’t fully understand it. The
> example that Zhu Zhu presented has a drawback of sending both tables to
> multiple nodes. What’s the benefit of using broadcast join over a hash join
> in such case? As far as I know, the biggest benefit of using broadcast join
> instead of hash join is that we can avoid sending the larger table over the
> network, because we can perform the join locally. In this example we are
> sending both of the tables to multiple nodes, which should defeat the
> purpose.
> >
> > Is it about implementing cross join or near cross joins in a distributed
> fashion?
> >
> >> if we introduce a new MulticastRecordWriter
> >
> > That’s one of the solutions. It might have a drawback of 3 class
> virtualisation problem (We have RecordWriter and BroadcastRecordWriter
> already). With up to two implementations, JVM is able to devirtualise the
> calls.
> >
> > Previously I was also thinking about just providing two different
> ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector`
> with plain `int` and based on that, RecordWriter could perform some magic
> (worst case scenario `instaceof` checks).
> >
> > Another solution might be to change `ChannelSelector` interface into an
> iterator.
> >
> > But let's discuss the details after we agree on implementing this.
> >
> > Piotrek
> >
> >> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email]> wrote:
> >>
> >>   Hi Piotr,
> >>
> >>        Thanks a lot for the suggestions!
> >>
> >>        The core motivation of this discussion is to implement a new
> iteration library on the DataStream, and it requires to insert special
> records in the stream to notify the progress of the iteration. The
> mechanism of such records is very similar to the current Watermark, and we
> meet the problem of sending normal records according to the partition
> (Rebalance, etc..) and also be able to broadcast the inserted progress
> records to all the connected records. I have read the notes in the google
> doc and I totally agree with that exploring the broadcast interface in
> RecordWriter in some way is able to solve this issue.
> >>
> >>       Regarding to `int[] ChannelSelector#selectChannels()`, I'm
> wondering if we introduce a new MulticastRecordWriter and left the current
> RecordWriter untouched, could we avoid the performance degradation ? Since
> with such a modification the normal RecordWriter does not need to iterate
> the return array by ChannelSelector, and the only difference will be
> returning an array instead of an integer, and accessing the first element
> of the returned array instead of reading the integer directly.
> >>
> >> Best,
> >> Yun
> >>
> >> ------------------------------------------------------------------
> >> From:Piotr Nowojski <[hidden email]>
> >> Send Time:2019 Aug. 23 (Fri.) 15:20
> >> To:dev <[hidden email]>
> >> Cc:Yun Gao <[hidden email]>
> >> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> Pattern
> >>
> >> Hi,
> >>
> >> Yun:
> >>
> >> Thanks for proposing the idea. I have checked the document and left
> couple of questions there, but it might be better to answer them here.
> >>
> >> What is the exact motivation and what problems do you want to solve? We
> have dropped multicast support from the network stack [1] for two reasons:
> >> 1. Performance
> >> 2. Code simplicity
> >>
> >> The proposal to re introduce `int[] ChannelSelector#selectChannels()`
> would revert those changes. At that time we were thinking about a way how
> to keep the multicast support on the network level, while keeping the
> performance and simplicity for non multicast cases and there are ways to
> achieve that. However they would add extra complexity to Flink, which it
> would be better to avoid.
> >>
> >> On the other hand, supporting dual pattern: standard partitioning or
> broadcasting is easy to do, as LatencyMarkers are doing exactly that. It
> would be just a matter of exposing this to the user in some way. So before
> we go any further, can you describe your use cases/motivation? Isn’t mix of
> standard partitioning and broadcasting enough? Do we need multicasting?
> >>
> >> Zhu:
> >>
> >> Could you rephrase your example? I didn’t quite understand it.
> >>
> >> Piotrek
> >>
> >> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> https://issues.apache.org/jira/browse/FLINK-10662>
> >>
> >> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
> [hidden email]>> wrote:
> >>
> >> Thanks Yun for starting this discussion.
> >> I think the multicasting can be very helpful in certain cases.
> >>
> >> I have received requirements from users that they want to do broadcast
> >> join, while the data set to broadcast is too large to fit in one task.
> >> Thus the requirement turned out to be to support cartesian product of 2
> >> data set(one of which can be infinite stream).
> >> For example, A(parallelism=2) broadcast join B(parallelism=2) in
> JobVertex
> >> C.
> >> The idea to is have 4 C subtasks to deal with different combinations of
> A/B
> >> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> >> This requires one record to be sent to multiple downstream subtasks, but
> >> not to all subtasks.
> >>
> >> With current interface this is not supported, as one record can only be
> >> sent to one subtask, or to all subtasks of a JobVertex.
> >> And the user had to split the broadcast data set manually to several
> >> different JobVertices, which is hard to maintain and extend.
> >>
> >> Thanks,
> >> Zhu Zhu
> >>
> >> Yun Gao <[hidden email] <mailto:
> [hidden email]>> 于2019年8月22日周四 下午8:42写道：
> >>
> >> Hi everyone,
> >>     In some scenarios we met a requirement that some operators want to
> >> send records to theirs downstream operators with an multicast
> communication
> >> pattern. In detail, for some records, the operators want to send them
> >> according to the partitioner (for example, Rebalance), and for some
> other
> >> records, the operators want to send them to all the connected operators
> and
> >> tasks. Such a communication pattern could be viewed as a kind of
> multicast:
> >> it does not broadcast every record, but some record will indeed be sent
> to
> >> multiple downstream operators.
> >>
> >> However, we found that this kind of communication pattern seems could
> not
> >> be implemented rightly if the operators have multiple consumers with
> >> different parallelism, using the customized partitioner. To solve the
> above
> >> problem, we propose to enhance the support for such kind of irregular
> >> communication pattern. We think there may be two options:
> >>
> >>    1. Support a kind of customized operator events, which share much
> >> similarity with Watermark, and these events can be broadcasted to the
> >> downstream operators separately.
> >>    2. Let the channel selector supports multicast, and also add the
> >> separate RecordWriter implementation to avoid impacting the performance
> of
> >> the channel selector that does not need multicast.
> >>
> >> The problem and options are detailed in
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> <
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >
> >>
> >> We are also wondering if there are other methods to implement this
> >> requirement with or without changing Runtime. Very thanks for any
> feedbacks
> >> !
> >>
> >>
> >> Best,
> >> Yun
> >>
> >>
> >>
> >>
> >
> >
>
>

[Piotr Nowojski-3]

Hi,

If the primary motivation is broadcasting (for the iterations) and we have no immediate need for multicast (cross join), I would prefer to first expose broadcast via the DataStream API and only later, once we finally need it, support multicast. As I wrote, multicast would be more challenging to implement, with more complicated runtime and API. And re-using multicast just to support broadcast doesn’t have much sense:

1. It’s a bit obfuscated. It’s easier to understand collectBroadcast(record) or broadcastEmit(record) compared to some multicast channel selector that just happens to return all of the channels.
2. There are performance benefits of explicitly calling `RecordWriter#broadcastEmit`.


On a different note, what would be the semantic of such broadcast emit on KeyedStream? Would it be supported? Or would we limit support only to the non-keyed streams?

Piotrek

> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
>
> Thanks Piotr,
>
> Users asked for this feature sometimes ago when they migrating batch jobs to Flink(Blink).
> It's not very urgent as they have taken some workarounds to solve it.(like partitioning data set to different job vertices)
> So it's fine to not make it top priority.
>
> Anyway, as a commonly known scenario, I think users can benefit from cross join sooner or later.
>
> Thanks,
> Zhu Zhu
>
> Piotr Nowojski <[hidden email] <mailto:[hidden email]>> 于2019年8月23日周五 下午6:19写道：
> Hi,
>
> Thanks for the answers :) Ok I understand the full picture now. +1 from my side on solving this issue somehow. But before we start discussing how to solve it one last control question:
>
> I guess this multicast is intended to be used in blink planner, right? Assuming that we implement the multicast support now, when would it be used by the blink? I would like to avoid a scenario, where we implement an unused feature and we keep maintaining it for a long period of time.
>
> Piotrek
>
> PS, try to include motivating examples, including concrete ones in the proposals/design docs, for example in the very first paragraph. Especially if it’s a commonly known feature like cross join :)
>
> > On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]> wrote:
> >
> >     Hi Piotr,
> >
> >        Thanks a lot for sharing the thoughts!
> >
> >        For the iteration, agree with that multicasting is not necessary. Exploring the broadcast interface to Output of the operators in some way should also solve this issue, and I think it should be even more convenient to have the broadcast method for the iteration.
> >
> >        Also thanks Zhu Zhu for the cross join case!
> >  Best,
> >   Yun
> >
> >
> >
> > ------------------------------------------------------------------
> > From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
> > Send Time:2019 Aug. 23 (Fri.) 17:25
> > To:dev <[hidden email] <mailto:[hidden email]>>
> > Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
> >
> > Hi Piotr,
> >
> > Yes you are right it's a distributed cross join requirement.
> > Broadcast join can help with cross join cases. But users cannot use it if the data set to join is too large to fit into one subtask.
> >
> > Sorry for left some details behind.
> >
> > Thanks,
> > Zhu Zhu
> > Piotr Nowojski <[hidden email] <mailto:[hidden email]>> 于2019年8月23日周五 下午4:57写道：
> > Hi Yun and Zhu Zhu,
> >
> > Thanks for the more detailed example Zhu Zhu.
> >
> > As far as I understand for the iterations example we do not need multicasting. Regarding the Join example, I don’t fully understand it. The example that Zhu Zhu presented has a drawback of sending both tables to multiple nodes. What’s the benefit of using broadcast join over a hash join in such case? As far as I know, the biggest benefit of using broadcast join instead of hash join is that we can avoid sending the larger table over the network, because we can perform the join locally. In this example we are sending both of the tables to multiple nodes, which should defeat the purpose.
> >
> > Is it about implementing cross join or near cross joins in a distributed fashion?
> >
> >> if we introduce a new MulticastRecordWriter
> >
> > That’s one of the solutions. It might have a drawback of 3 class virtualisation problem (We have RecordWriter and BroadcastRecordWriter already). With up to two implementations, JVM is able to devirtualise the calls.
> >
> > Previously I was also thinking about just providing two different ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector` with plain `int` and based on that, RecordWriter could perform some magic (worst case scenario `instaceof` checks).
> >
> > Another solution might be to change `ChannelSelector` interface into an iterator.
> >
> > But let's discuss the details after we agree on implementing this.
> >
> > Piotrek
> >
> >> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:[hidden email]>> wrote:
> >>
> >>   Hi Piotr,
> >>
> >>        Thanks a lot for the suggestions!
> >>
> >>        The core motivation of this discussion is to implement a new iteration library on the DataStream, and it requires to insert special records in the stream to notify the progress of the iteration. The mechanism of such records is very similar to the current Watermark, and we meet the problem of sending normal records according to the partition (Rebalance, etc..) and also be able to broadcast the inserted progress records to all the connected records. I have read the notes in the google doc and I totally agree with that exploring the broadcast interface in RecordWriter in some way is able to solve this issue.
> >>
> >>       Regarding to `int[] ChannelSelector#selectChannels()`, I'm wondering if we introduce a new MulticastRecordWriter and left the current RecordWriter untouched, could we avoid the performance degradation ? Since with such a modification the normal RecordWriter does not need to iterate the return array by ChannelSelector, and the only difference will be returning an array instead of an integer, and accessing the first element of the returned array instead of reading the integer directly.
> >>
> >> Best,
> >> Yun
> >>
> >> ------------------------------------------------------------------
> >> From:Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> >> Send Time:2019 Aug. 23 (Fri.) 15:20
> >> To:dev <[hidden email] <mailto:[hidden email]>>
> >> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> >> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
> >>
> >> Hi,
> >>
> >> Yun:
> >>
> >> Thanks for proposing the idea. I have checked the document and left couple of questions there, but it might be better to answer them here.
> >>
> >> What is the exact motivation and what problems do you want to solve? We have dropped multicast support from the network stack [1] for two reasons:
> >> 1. Performance
> >> 2. Code simplicity
> >>
> >> The proposal to re introduce `int[] ChannelSelector#selectChannels()` would revert those changes. At that time we were thinking about a way how to keep the multicast support on the network level, while keeping the performance and simplicity for non multicast cases and there are ways to achieve that. However they would add extra complexity to Flink, which it would be better to avoid.
> >>
> >> On the other hand, supporting dual pattern: standard partitioning or broadcasting is easy to do, as LatencyMarkers are doing exactly that. It would be just a matter of exposing this to the user in some way. So before we go any further, can you describe your use cases/motivation? Isn’t mix of standard partitioning and broadcasting enough? Do we need multicasting?
> >>
> >> Zhu:
> >>
> >> Could you rephrase your example? I didn’t quite understand it.
> >>
> >> Piotrek
> >>
> >> [1] https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662> <https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662>>
> >>
> >> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
> >>
> >> Thanks Yun for starting this discussion.
> >> I think the multicasting can be very helpful in certain cases.
> >>
> >> I have received requirements from users that they want to do broadcast
> >> join, while the data set to broadcast is too large to fit in one task.
> >> Thus the requirement turned out to be to support cartesian product of 2
> >> data set(one of which can be infinite stream).
> >> For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
> >> C.
> >> The idea to is have 4 C subtasks to deal with different combinations of A/B
> >> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> >> This requires one record to be sent to multiple downstream subtasks, but
> >> not to all subtasks.
> >>
> >> With current interface this is not supported, as one record can only be
> >> sent to one subtask, or to all subtasks of a JobVertex.
> >> And the user had to split the broadcast data set manually to several
> >> different JobVertices, which is hard to maintain and extend.
> >>
> >> Thanks,
> >> Zhu Zhu
> >>
> >> Yun Gao <[hidden email] <mailto:[hidden email] <mailto:[hidden email]>>> 于2019年8月22日周四 下午8:42写道：
> >>
> >> Hi everyone,
> >>     In some scenarios we met a requirement that some operators want to
> >> send records to theirs downstream operators with an multicast communication
> >> pattern. In detail, for some records, the operators want to send them
> >> according to the partitioner (for example, Rebalance), and for some other
> >> records, the operators want to send them to all the connected operators and
> >> tasks. Such a communication pattern could be viewed as a kind of multicast:
> >> it does not broadcast every record, but some record will indeed be sent to
> >> multiple downstream operators.
> >>
> >> However, we found that this kind of communication pattern seems could not
> >> be implemented rightly if the operators have multiple consumers with
> >> different parallelism, using the customized partitioner. To solve the above
> >> problem, we propose to enhance the support for such kind of irregular
> >> communication pattern. We think there may be two options:
> >>
> >>    1. Support a kind of customized operator events, which share much
> >> similarity with Watermark, and these events can be broadcasted to the
> >> downstream operators separately.
> >>    2. Let the channel selector supports multicast, and also add the
> >> separate RecordWriter implementation to avoid impacting the performance of
> >> the channel selector that does not need multicast.
> >>
> >> The problem and options are detailed in
> >> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing> <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing>>
> >>
> >> We are also wondering if there are other methods to implement this
> >> requirement with or without changing Runtime. Very thanks for any feedbacks
> >> !
> >>
> >>
> >> Best,
> >> Yun
> >>
> >>
> >>
> >>
> >
> >
>

[Yun Gao]

     Hi Piotr,

      Very thanks for the suggestions!  

     Totally agree with that we could first focus on the broadcast scenarios and exposing the broadcastEmit method first considering the semantics and performance.

     For the keyed stream, I also agree with that broadcasting keyed records to all the tasks may be confused considering the semantics of keyed partitioner. However, in the iteration case supporting broadcast over keyed partitioner should be required since users may create any subgraph for the iteration body, including the operators with key. I think a possible solution to this issue is to introduce another data type for 'broadcastEmit'. For example, for an operator Operator<T>, it may broadcast emit another type E instead of T, and the transmitting E will bypass the partitioner and setting keyed context. This should result in the design to introduce customized operator event (option 1 in the document). The cost of this method is that we need to introduce a new type of StreamElement and new interface for this type, but it should be suitable for both keyed or non-keyed partitioner.

Best,
Yun



------------------------------------------------------------------
From:Piotr Nowojski <[hidden email]>
Send Time:2019 Aug. 23 (Fri.) 22:29
To:Zhu Zhu <[hidden email]>
Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern

Hi,

If the primary motivation is broadcasting (for the iterations) and we have no immediate need for multicast (cross join), I would prefer to first expose broadcast via the DataStream API and only later, once we finally need it, support multicast. As I wrote, multicast would be more challenging to implement, with more complicated runtime and API. And re-using multicast just to support broadcast doesn’t have much sense:

1. It’s a bit obfuscated. It’s easier to understand collectBroadcast(record) or broadcastEmit(record) compared to some multicast channel selector that just happens to return all of the channels.
2. There are performance benefits of explicitly calling `RecordWriter#broadcastEmit`.


On a different note, what would be the semantic of such broadcast emit on KeyedStream? Would it be supported? Or would we limit support only to the non-keyed streams?

Piotrek

> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
>
> Thanks Piotr,
>
> Users asked for this feature sometimes ago when they migrating batch jobs to Flink(Blink).
> It's not very urgent as they have taken some workarounds to solve it.(like partitioning data set to different job vertices)
> So it's fine to not make it top priority.
>
> Anyway, as a commonly known scenario, I think users can benefit from cross join sooner or later.
>
> Thanks,
> Zhu Zhu
>
> Piotr Nowojski <[hidden email] <mailto:[hidden email]>> 于2019年8月23日周五 下午6:19写道：
> Hi,
>
> Thanks for the answers :) Ok I understand the full picture now. +1 from my side on solving this issue somehow. But before we start discussing how to solve it one last control question:
>
> I guess this multicast is intended to be used in blink planner, right? Assuming that we implement the multicast support now, when would it be used by the blink? I would like to avoid a scenario, where we implement an unused feature and we keep maintaining it for a long period of time.
>
> Piotrek
>
> PS, try to include motivating examples, including concrete ones in the proposals/design docs, for example in the very first paragraph. Especially if it’s a commonly known feature like cross join :)
>
> > On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]> wrote:
> >
> >     Hi Piotr,
> >
> >        Thanks a lot for sharing the thoughts!
> >
> >        For the iteration, agree with that multicasting is not necessary. Exploring the broadcast interface to Output of the operators in some way should also solve this issue, and I think it should be even more convenient to have the broadcast method for the iteration.
> >
> >        Also thanks Zhu Zhu for the cross join case!
> >  Best,
> >   Yun
> >
> >
> >
> > ------------------------------------------------------------------
> > From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
> > Send Time:2019 Aug. 23 (Fri.) 17:25
> > To:dev <[hidden email] <mailto:[hidden email]>>
> > Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
> >
> > Hi Piotr,
> >
> > Yes you are right it's a distributed cross join requirement.
> > Broadcast join can help with cross join cases. But users cannot use it if the data set to join is too large to fit into one subtask.
> >
> > Sorry for left some details behind.
> >
> > Thanks,
> > Zhu Zhu
> > Piotr Nowojski <[hidden email] <mailto:[hidden email]>> 于2019年8月23日周五 下午4:57写道：
> > Hi Yun and Zhu Zhu,
> >
> > Thanks for the more detailed example Zhu Zhu.
> >
> > As far as I understand for the iterations example we do not need multicasting. Regarding the Join example, I don’t fully understand it. The example that Zhu Zhu presented has a drawback of sending both tables to multiple nodes. What’s the benefit of using broadcast join over a hash join in such case? As far as I know, the biggest benefit of using broadcast join instead of hash join is that we can avoid sending the larger table over the network, because we can perform the join locally. In this example we are sending both of the tables to multiple nodes, which should defeat the purpose.
> >
> > Is it about implementing cross join or near cross joins in a distributed fashion?
> >
> >> if we introduce a new MulticastRecordWriter
> >
> > That’s one of the solutions. It might have a drawback of 3 class virtualisation problem (We have RecordWriter and BroadcastRecordWriter already). With up to two implementations, JVM is able to devirtualise the calls.
> >
> > Previously I was also thinking about just providing two different ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector` with plain `int` and based on that, RecordWriter could perform some magic (worst case scenario `instaceof` checks).
> >
> > Another solution might be to change `ChannelSelector` interface into an iterator.
> >
> > But let's discuss the details after we agree on implementing this.
> >
> > Piotrek
> >
> >> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:[hidden email]>> wrote:
> >>
> >>   Hi Piotr,
> >>
> >>        Thanks a lot for the suggestions!
> >>
> >>        The core motivation of this discussion is to implement a new iteration library on the DataStream, and it requires to insert special records in the stream to notify the progress of the iteration. The mechanism of such records is very similar to the current Watermark, and we meet the problem of sending normal records according to the partition (Rebalance, etc..) and also be able to broadcast the inserted progress records to all the connected records. I have read the notes in the google doc and I totally agree with that exploring the broadcast interface in RecordWriter in some way is able to solve this issue.
> >>
> >>       Regarding to `int[] ChannelSelector#selectChannels()`, I'm wondering if we introduce a new MulticastRecordWriter and left the current RecordWriter untouched, could we avoid the performance degradation ? Since with such a modification the normal RecordWriter does not need to iterate the return array by ChannelSelector, and the only difference will be returning an array instead of an integer, and accessing the first element of the returned array instead of reading the integer directly.
> >>
> >> Best,
> >> Yun
> >>
> >> ------------------------------------------------------------------
> >> From:Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> >> Send Time:2019 Aug. 23 (Fri.) 15:20
> >> To:dev <[hidden email] <mailto:[hidden email]>>
> >> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> >> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
> >>
> >> Hi,
> >>
> >> Yun:
> >>
> >> Thanks for proposing the idea. I have checked the document and left couple of questions there, but it might be better to answer them here.
> >>
> >> What is the exact motivation and what problems do you want to solve? We have dropped multicast support from the network stack [1] for two reasons:
> >> 1. Performance
> >> 2. Code simplicity
> >>
> >> The proposal to re introduce `int[] ChannelSelector#selectChannels()` would revert those changes. At that time we were thinking about a way how to keep the multicast support on the network level, while keeping the performance and simplicity for non multicast cases and there are ways to achieve that. However they would add extra complexity to Flink, which it would be better to avoid.
> >>
> >> On the other hand, supporting dual pattern: standard partitioning or broadcasting is easy to do, as LatencyMarkers are doing exactly that. It would be just a matter of exposing this to the user in some way. So before we go any further, can you describe your use cases/motivation? Isn’t mix of standard partitioning and broadcasting enough? Do we need multicasting?
> >>
> >> Zhu:
> >>
> >> Could you rephrase your example? I didn’t quite understand it.
> >>
> >> Piotrek
> >>
> >> [1] https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662> <https://issues.apache.org/jira/browse/FLINK-10662 <https://issues.apache.org/jira/browse/FLINK-10662>>
> >>
> >> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:[hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>> wrote:
> >>
> >> Thanks Yun for starting this discussion.
> >> I think the multicasting can be very helpful in certain cases.
> >>
> >> I have received requirements from users that they want to do broadcast
> >> join, while the data set to broadcast is too large to fit in one task.
> >> Thus the requirement turned out to be to support cartesian product of 2
> >> data set(one of which can be infinite stream).
> >> For example, A(parallelism=2) broadcast join B(parallelism=2) in JobVertex
> >> C.
> >> The idea to is have 4 C subtasks to deal with different combinations of A/B
> >> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> >> This requires one record to be sent to multiple downstream subtasks, but
> >> not to all subtasks.
> >>
> >> With current interface this is not supported, as one record can only be
> >> sent to one subtask, or to all subtasks of a JobVertex.
> >> And the user had to split the broadcast data set manually to several
> >> different JobVertices, which is hard to maintain and extend.
> >>
> >> Thanks,
> >> Zhu Zhu
> >>
> >> Yun Gao <[hidden email] <mailto:[hidden email] <mailto:[hidden email]>>> 于2019年8月22日周四 下午8:42写道：
> >>
> >> Hi everyone,
> >>     In some scenarios we met a requirement that some operators want to
> >> send records to theirs downstream operators with an multicast communication
> >> pattern. In detail, for some records, the operators want to send them
> >> according to the partitioner (for example, Rebalance), and for some other
> >> records, the operators want to send them to all the connected operators and
> >> tasks. Such a communication pattern could be viewed as a kind of multicast:
> >> it does not broadcast every record, but some record will indeed be sent to
> >> multiple downstream operators.
> >>
> >> However, we found that this kind of communication pattern seems could not
> >> be implemented rightly if the operators have multiple consumers with
> >> different parallelism, using the customized partitioner. To solve the above
> >> problem, we propose to enhance the support for such kind of irregular
> >> communication pattern. We think there may be two options:
> >>
> >>    1. Support a kind of customized operator events, which share much
> >> similarity with Watermark, and these events can be broadcasted to the
> >> downstream operators separately.
> >>    2. Let the channel selector supports multicast, and also add the
> >> separate RecordWriter implementation to avoid impacting the performance of
> >> the channel selector that does not need multicast.
> >>
> >> The problem and options are detailed in
> >> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing> <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing <https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing>>
> >>
> >> We are also wondering if there are other methods to implement this
> >> requirement with or without changing Runtime. Very thanks for any feedbacks
> >> !
> >>
> >>
> >> Best,
> >> Yun
> >>
> >>
> >>
> >>
> >
> >
>

[Zhu Zhu]

Hi Piotr,

Thanks for the explanation.
Agreed that the broadcastEmit(record) is a better choice for broadcasting
for the iterations.
As broadcasting for the iterations is the first motivation, let's support
it first.

Thanks,
Zhu Zhu

Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：

>      Hi Piotr,
>
>       Very thanks for the suggestions!
>
>      Totally agree with that we could first focus on the broadcast
> scenarios and exposing the broadcastEmit method first considering the
> semantics and performance.
>
>      For the keyed stream, I also agree with that broadcasting keyed
> records to all the tasks may be confused considering the semantics of keyed
> partitioner. However, in the iteration case supporting broadcast over keyed
> partitioner should be required since users may create any subgraph for the
> iteration body, including the operators with key. I think a possible
> solution to this issue is to introduce another data type for
> 'broadcastEmit'. For example, for an operator Operator<T>, it may broadcast
> emit another type E instead of T, and the transmitting E will bypass the
> partitioner and setting keyed context. This should result in the design to
> introduce customized operator event (option 1 in the document). The cost of
> this method is that we need to introduce a new type of StreamElement and
> new interface for this type, but it should be suitable for both keyed or
> non-keyed partitioner.
>
> Best,
> Yun
>
>
>
> ------------------------------------------------------------------
> From:Piotr Nowojski <[hidden email]>
> Send Time:2019 Aug. 23 (Fri.) 22:29
> To:Zhu Zhu <[hidden email]>
> Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
>
> Hi,
>
> If the primary motivation is broadcasting (for the iterations) and we have
> no immediate need for multicast (cross join), I would prefer to first
> expose broadcast via the DataStream API and only later, once we finally
> need it, support multicast. As I wrote, multicast would be more challenging
> to implement, with more complicated runtime and API. And re-using multicast
> just to support broadcast doesn’t have much sense:
>
> 1. It’s a bit obfuscated. It’s easier to understand
> collectBroadcast(record) or broadcastEmit(record) compared to some
> multicast channel selector that just happens to return all of the channels.
> 2. There are performance benefits of explicitly calling
> `RecordWriter#broadcastEmit`.
>
>
> On a different note, what would be the semantic of such broadcast emit on
> KeyedStream? Would it be supported? Or would we limit support only to the
> non-keyed streams?
>
> Piotrek
>
> > On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
> >
> > Thanks Piotr,
> >
> > Users asked for this feature sometimes ago when they migrating batch
> jobs to Flink(Blink).
> > It's not very urgent as they have taken some workarounds to solve
> it.(like partitioning data set to different job vertices)
> > So it's fine to not make it top priority.
> >
> > Anyway, as a commonly known scenario, I think users can benefit from
> cross join sooner or later.
> >
> > Thanks,
> > Zhu Zhu
> >
> > Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> 于2019年8月23日周五 下午6:19写道：
> > Hi,
> >
> > Thanks for the answers :) Ok I understand the full picture now. +1 from
> my side on solving this issue somehow. But before we start discussing how
> to solve it one last control question:
> >
> > I guess this multicast is intended to be used in blink planner, right?
> Assuming that we implement the multicast support now, when would it be used
> by the blink? I would like to avoid a scenario, where we implement an
> unused feature and we keep maintaining it for a long period of time.
> >
> > Piotrek
> >
> > PS, try to include motivating examples, including concrete ones in the
> proposals/design docs, for example in the very first paragraph. Especially
> if it’s a commonly known feature like cross join :)
> >
> > > On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
> wrote:
> > >
> > >     Hi Piotr,
> > >
> > >        Thanks a lot for sharing the thoughts!
> > >
> > >        For the iteration, agree with that multicasting is not
> necessary. Exploring the broadcast interface to Output of the operators in
> some way should also solve this issue, and I think it should be even more
> convenient to have the broadcast method for the iteration.
> > >
> > >        Also thanks Zhu Zhu for the cross join case!
> > >  Best,
> > >   Yun
> > >
> > >
> > >
> > > ------------------------------------------------------------------
> > > From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
> > > Send Time:2019 Aug. 23 (Fri.) 17:25
> > > To:dev <[hidden email] <mailto:[hidden email]>>
> > > Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > > Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> Pattern
> > >
> > > Hi Piotr,
> > >
> > > Yes you are right it's a distributed cross join requirement.
> > > Broadcast join can help with cross join cases. But users cannot use it
> if the data set to join is too large to fit into one subtask.
> > >
> > > Sorry for left some details behind.
> > >
> > > Thanks,
> > > Zhu Zhu
> > > Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> 于2019年8月23日周五 下午4:57写道：
> > > Hi Yun and Zhu Zhu,
> > >
> > > Thanks for the more detailed example Zhu Zhu.
> > >
> > > As far as I understand for the iterations example we do not need
> multicasting. Regarding the Join example, I don’t fully understand it. The
> example that Zhu Zhu presented has a drawback of sending both tables to
> multiple nodes. What’s the benefit of using broadcast join over a hash join
> in such case? As far as I know, the biggest benefit of using broadcast join
> instead of hash join is that we can avoid sending the larger table over the
> network, because we can perform the join locally. In this example we are
> sending both of the tables to multiple nodes, which should defeat the
> purpose.
> > >
> > > Is it about implementing cross join or near cross joins in a
> distributed fashion?
> > >
> > >> if we introduce a new MulticastRecordWriter
> > >
> > > That’s one of the solutions. It might have a drawback of 3 class
> virtualisation problem (We have RecordWriter and BroadcastRecordWriter
> already). With up to two implementations, JVM is able to devirtualise the
> calls.
> > >
> > > Previously I was also thinking about just providing two different
> ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector`
> with plain `int` and based on that, RecordWriter could perform some magic
> (worst case scenario `instaceof` checks).
> > >
> > > Another solution might be to change `ChannelSelector` interface into
> an iterator.
> > >
> > > But let's discuss the details after we agree on implementing this.
> > >
> > > Piotrek
> > >
> > >> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
> [hidden email]>> wrote:
> > >>
> > >>   Hi Piotr,
> > >>
> > >>        Thanks a lot for the suggestions!
> > >>
> > >>        The core motivation of this discussion is to implement a new
> iteration library on the DataStream, and it requires to insert special
> records in the stream to notify the progress of the iteration. The
> mechanism of such records is very similar to the current Watermark, and we
> meet the problem of sending normal records according to the partition
> (Rebalance, etc..) and also be able to broadcast the inserted progress
> records to all the connected records. I have read the notes in the google
> doc and I totally agree with that exploring the broadcast interface in
> RecordWriter in some way is able to solve this issue.
> > >>
> > >>       Regarding to `int[] ChannelSelector#selectChannels()`, I'm
> wondering if we introduce a new MulticastRecordWriter and left the current
> RecordWriter untouched, could we avoid the performance degradation ? Since
> with such a modification the normal RecordWriter does not need to iterate
> the return array by ChannelSelector, and the only difference will be
> returning an array instead of an integer, and accessing the first element
> of the returned array instead of reading the integer directly.
> > >>
> > >> Best,
> > >> Yun
> > >>
> > >> ------------------------------------------------------------------
> > >> From:Piotr Nowojski <[hidden email] <mailto:[hidden email]
> >>
> > >> Send Time:2019 Aug. 23 (Fri.) 15:20
> > >> To:dev <[hidden email] <mailto:[hidden email]>>
> > >> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > >> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> Pattern
> > >>
> > >> Hi,
> > >>
> > >> Yun:
> > >>
> > >> Thanks for proposing the idea. I have checked the document and left
> couple of questions there, but it might be better to answer them here.
> > >>
> > >> What is the exact motivation and what problems do you want to solve?
> We have dropped multicast support from the network stack [1] for two
> reasons:
> > >> 1. Performance
> > >> 2. Code simplicity
> > >>
> > >> The proposal to re introduce `int[] ChannelSelector#selectChannels()`
> would revert those changes. At that time we were thinking about a way how
> to keep the multicast support on the network level, while keeping the
> performance and simplicity for non multicast cases and there are ways to
> achieve that. However they would add extra complexity to Flink, which it
> would be better to avoid.
> > >>
> > >> On the other hand, supporting dual pattern: standard partitioning or
> broadcasting is easy to do, as LatencyMarkers are doing exactly that. It
> would be just a matter of exposing this to the user in some way. So before
> we go any further, can you describe your use cases/motivation? Isn’t mix of
> standard partitioning and broadcasting enough? Do we need multicasting?
> > >>
> > >> Zhu:
> > >>
> > >> Could you rephrase your example? I didn’t quite understand it.
> > >>
> > >> Piotrek
> > >>
> > >> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> https://issues.apache.org/jira/browse/FLINK-10662> <
> https://issues.apache.org/jira/browse/FLINK-10662 <
> https://issues.apache.org/jira/browse/FLINK-10662>>
> > >>
> > >> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
> [hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>
> wrote:
> > >>
> > >> Thanks Yun for starting this discussion.
> > >> I think the multicasting can be very helpful in certain cases.
> > >>
> > >> I have received requirements from users that they want to do broadcast
> > >> join, while the data set to broadcast is too large to fit in one task.
> > >> Thus the requirement turned out to be to support cartesian product of
> 2
> > >> data set(one of which can be infinite stream).
> > >> For example, A(parallelism=2) broadcast join B(parallelism=2) in
> JobVertex
> > >> C.
> > >> The idea to is have 4 C subtasks to deal with different combinations
> of A/B
> > >> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> > >> This requires one record to be sent to multiple downstream subtasks,
> but
> > >> not to all subtasks.
> > >>
> > >> With current interface this is not supported, as one record can only
> be
> > >> sent to one subtask, or to all subtasks of a JobVertex.
> > >> And the user had to split the broadcast data set manually to several
> > >> different JobVertices, which is hard to maintain and extend.
> > >>
> > >> Thanks,
> > >> Zhu Zhu
> > >>
> > >> Yun Gao <[hidden email] <mailto:
> [hidden email] <mailto:[hidden email]>>>
> 于2019年8月22日周四 下午8:42写道：
> > >>
> > >> Hi everyone,
> > >>     In some scenarios we met a requirement that some operators want to
> > >> send records to theirs downstream operators with an multicast
> communication
> > >> pattern. In detail, for some records, the operators want to send them
> > >> according to the partitioner (for example, Rebalance), and for some
> other
> > >> records, the operators want to send them to all the connected
> operators and
> > >> tasks. Such a communication pattern could be viewed as a kind of
> multicast:
> > >> it does not broadcast every record, but some record will indeed be
> sent to
> > >> multiple downstream operators.
> > >>
> > >> However, we found that this kind of communication pattern seems could
> not
> > >> be implemented rightly if the operators have multiple consumers with
> > >> different parallelism, using the customized partitioner. To solve the
> above
> > >> problem, we propose to enhance the support for such kind of irregular
> > >> communication pattern. We think there may be two options:
> > >>
> > >>    1. Support a kind of customized operator events, which share much
> > >> similarity with Watermark, and these events can be broadcasted to the
> > >> downstream operators separately.
> > >>    2. Let the channel selector supports multicast, and also add the
> > >> separate RecordWriter implementation to avoid impacting the
> performance of
> > >> the channel selector that does not need multicast.
> > >>
> > >> The problem and options are detailed in
> > >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> <
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing>
> <
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> <
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >>
> > >>
> > >> We are also wondering if there are other methods to implement this
> > >> requirement with or without changing Runtime. Very thanks for any
> feedbacks
> > >> !
> > >>
> > >>
> > >> Best,
> > >> Yun
> > >>
> > >>
> > >>
> > >>
> > >
> > >
> >
>
>
>

[xingcanc]

Hi all,

Sorry for joining this thread late. Basically, I think enabling multicast pattern could be the right direction, but more detailed implementation policies need to be discussed.

Two years ago, I filed an issue [1] about the multicast API. However, due to some reasons, it was laid aside. After that, when I tried to cherry-pick the change for experimental use, I found the return type of `selectChannels()` method had changed from `int[]` to `int`, which makes the old implementation not work anymore.

From my side, the multicast has always been used for theta-join. As far as I know, it’s an essential requirement for some sophisticated joining algorithms. Until now, the Flink non-equi joins can still only be executed single-threaded. If we'd like to make some improvements on this, we should first take some measures to support multicast pattern.

Best,
Xingcan

[1] https://issues.apache.org/jira/browse/FLINK-6936

> On Aug 24, 2019, at 5:54 AM, Zhu Zhu <[hidden email]> wrote:
>
> Hi Piotr,
>
> Thanks for the explanation.
> Agreed that the broadcastEmit(record) is a better choice for broadcasting
> for the iterations.
> As broadcasting for the iterations is the first motivation, let's support
> it first.
>
> Thanks,
> Zhu Zhu
>
> Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：
>
>>     Hi Piotr,
>>
>>      Very thanks for the suggestions!
>>
>>     Totally agree with that we could first focus on the broadcast
>> scenarios and exposing the broadcastEmit method first considering the
>> semantics and performance.
>>
>>     For the keyed stream, I also agree with that broadcasting keyed
>> records to all the tasks may be confused considering the semantics of keyed
>> partitioner. However, in the iteration case supporting broadcast over keyed
>> partitioner should be required since users may create any subgraph for the
>> iteration body, including the operators with key. I think a possible
>> solution to this issue is to introduce another data type for
>> 'broadcastEmit'. For example, for an operator Operator<T>, it may broadcast
>> emit another type E instead of T, and the transmitting E will bypass the
>> partitioner and setting keyed context. This should result in the design to
>> introduce customized operator event (option 1 in the document). The cost of
>> this method is that we need to introduce a new type of StreamElement and
>> new interface for this type, but it should be suitable for both keyed or
>> non-keyed partitioner.
>>
>> Best,
>> Yun
>>
>>
>>
>> ------------------------------------------------------------------
>> From:Piotr Nowojski <[hidden email]>
>> Send Time:2019 Aug. 23 (Fri.) 22:29
>> To:Zhu Zhu <[hidden email]>
>> Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern
>>
>> Hi,
>>
>> If the primary motivation is broadcasting (for the iterations) and we have
>> no immediate need for multicast (cross join), I would prefer to first
>> expose broadcast via the DataStream API and only later, once we finally
>> need it, support multicast. As I wrote, multicast would be more challenging
>> to implement, with more complicated runtime and API. And re-using multicast
>> just to support broadcast doesn’t have much sense:
>>
>> 1. It’s a bit obfuscated. It’s easier to understand
>> collectBroadcast(record) or broadcastEmit(record) compared to some
>> multicast channel selector that just happens to return all of the channels.
>> 2. There are performance benefits of explicitly calling
>> `RecordWriter#broadcastEmit`.
>>
>>
>> On a different note, what would be the semantic of such broadcast emit on
>> KeyedStream? Would it be supported? Or would we limit support only to the
>> non-keyed streams?
>>
>> Piotrek
>>
>>> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
>>>
>>> Thanks Piotr,
>>>
>>> Users asked for this feature sometimes ago when they migrating batch
>> jobs to Flink(Blink).
>>> It's not very urgent as they have taken some workarounds to solve
>> it.(like partitioning data set to different job vertices)
>>> So it's fine to not make it top priority.
>>>
>>> Anyway, as a commonly known scenario, I think users can benefit from
>> cross join sooner or later.
>>>
>>> Thanks,
>>> Zhu Zhu
>>>
>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
>> 于2019年8月23日周五 下午6:19写道：
>>> Hi,
>>>
>>> Thanks for the answers :) Ok I understand the full picture now. +1 from
>> my side on solving this issue somehow. But before we start discussing how
>> to solve it one last control question:
>>>
>>> I guess this multicast is intended to be used in blink planner, right?
>> Assuming that we implement the multicast support now, when would it be used
>> by the blink? I would like to avoid a scenario, where we implement an
>> unused feature and we keep maintaining it for a long period of time.
>>>
>>> Piotrek
>>>
>>> PS, try to include motivating examples, including concrete ones in the
>> proposals/design docs, for example in the very first paragraph. Especially
>> if it’s a commonly known feature like cross join :)
>>>
>>>> On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
>> wrote:
>>>>
>>>>    Hi Piotr,
>>>>
>>>>       Thanks a lot for sharing the thoughts!
>>>>
>>>>       For the iteration, agree with that multicasting is not
>> necessary. Exploring the broadcast interface to Output of the operators in
>> some way should also solve this issue, and I think it should be even more
>> convenient to have the broadcast method for the iteration.
>>>>
>>>>       Also thanks Zhu Zhu for the cross join case!
>>>> Best,
>>>>  Yun
>>>>
>>>>
>>>>
>>>> ------------------------------------------------------------------
>>>> From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
>>>> Send Time:2019 Aug. 23 (Fri.) 17:25
>>>> To:dev <[hidden email] <mailto:[hidden email]>>
>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
>> Pattern
>>>>
>>>> Hi Piotr,
>>>>
>>>> Yes you are right it's a distributed cross join requirement.
>>>> Broadcast join can help with cross join cases. But users cannot use it
>> if the data set to join is too large to fit into one subtask.
>>>>
>>>> Sorry for left some details behind.
>>>>
>>>> Thanks,
>>>> Zhu Zhu
>>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
>> 于2019年8月23日周五 下午4:57写道：
>>>> Hi Yun and Zhu Zhu,
>>>>
>>>> Thanks for the more detailed example Zhu Zhu.
>>>>
>>>> As far as I understand for the iterations example we do not need
>> multicasting. Regarding the Join example, I don’t fully understand it. The
>> example that Zhu Zhu presented has a drawback of sending both tables to
>> multiple nodes. What’s the benefit of using broadcast join over a hash join
>> in such case? As far as I know, the biggest benefit of using broadcast join
>> instead of hash join is that we can avoid sending the larger table over the
>> network, because we can perform the join locally. In this example we are
>> sending both of the tables to multiple nodes, which should defeat the
>> purpose.
>>>>
>>>> Is it about implementing cross join or near cross joins in a
>> distributed fashion?
>>>>
>>>>> if we introduce a new MulticastRecordWriter
>>>>
>>>> That’s one of the solutions. It might have a drawback of 3 class
>> virtualisation problem (We have RecordWriter and BroadcastRecordWriter
>> already). With up to two implementations, JVM is able to devirtualise the
>> calls.
>>>>
>>>> Previously I was also thinking about just providing two different
>> ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector`
>> with plain `int` and based on that, RecordWriter could perform some magic
>> (worst case scenario `instaceof` checks).
>>>>
>>>> Another solution might be to change `ChannelSelector` interface into
>> an iterator.
>>>>
>>>> But let's discuss the details after we agree on implementing this.
>>>>
>>>> Piotrek
>>>>
>>>>> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
>> [hidden email]>> wrote:
>>>>>
>>>>>  Hi Piotr,
>>>>>
>>>>>       Thanks a lot for the suggestions!
>>>>>
>>>>>       The core motivation of this discussion is to implement a new
>> iteration library on the DataStream, and it requires to insert special
>> records in the stream to notify the progress of the iteration. The
>> mechanism of such records is very similar to the current Watermark, and we
>> meet the problem of sending normal records according to the partition
>> (Rebalance, etc..) and also be able to broadcast the inserted progress
>> records to all the connected records. I have read the notes in the google
>> doc and I totally agree with that exploring the broadcast interface in
>> RecordWriter in some way is able to solve this issue.
>>>>>
>>>>>      Regarding to `int[] ChannelSelector#selectChannels()`, I'm
>> wondering if we introduce a new MulticastRecordWriter and left the current
>> RecordWriter untouched, could we avoid the performance degradation ? Since
>> with such a modification the normal RecordWriter does not need to iterate
>> the return array by ChannelSelector, and the only difference will be
>> returning an array instead of an integer, and accessing the first element
>> of the returned array instead of reading the integer directly.
>>>>>
>>>>> Best,
>>>>> Yun
>>>>>
>>>>> ------------------------------------------------------------------
>>>>> From:Piotr Nowojski <[hidden email] <mailto:[hidden email]
>>>>
>>>>> Send Time:2019 Aug. 23 (Fri.) 15:20
>>>>> To:dev <[hidden email] <mailto:[hidden email]>>
>>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
>>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
>> Pattern
>>>>>
>>>>> Hi,
>>>>>
>>>>> Yun:
>>>>>
>>>>> Thanks for proposing the idea. I have checked the document and left
>> couple of questions there, but it might be better to answer them here.
>>>>>
>>>>> What is the exact motivation and what problems do you want to solve?
>> We have dropped multicast support from the network stack [1] for two
>> reasons:
>>>>> 1. Performance
>>>>> 2. Code simplicity
>>>>>
>>>>> The proposal to re introduce `int[] ChannelSelector#selectChannels()`
>> would revert those changes. At that time we were thinking about a way how
>> to keep the multicast support on the network level, while keeping the
>> performance and simplicity for non multicast cases and there are ways to
>> achieve that. However they would add extra complexity to Flink, which it
>> would be better to avoid.
>>>>>
>>>>> On the other hand, supporting dual pattern: standard partitioning or
>> broadcasting is easy to do, as LatencyMarkers are doing exactly that. It
>> would be just a matter of exposing this to the user in some way. So before
>> we go any further, can you describe your use cases/motivation? Isn’t mix of
>> standard partitioning and broadcasting enough? Do we need multicasting?
>>>>>
>>>>> Zhu:
>>>>>
>>>>> Could you rephrase your example? I didn’t quite understand it.
>>>>>
>>>>> Piotrek
>>>>>
>>>>> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
>> https://issues.apache.org/jira/browse/FLINK-10662> <
>> https://issues.apache.org/jira/browse/FLINK-10662 <
>> https://issues.apache.org/jira/browse/FLINK-10662>>
>>>>>
>>>>> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
>> [hidden email]> <mailto:[hidden email] <mailto:[hidden email]>>>
>> wrote:
>>>>>
>>>>> Thanks Yun for starting this discussion.
>>>>> I think the multicasting can be very helpful in certain cases.
>>>>>
>>>>> I have received requirements from users that they want to do broadcast
>>>>> join, while the data set to broadcast is too large to fit in one task.
>>>>> Thus the requirement turned out to be to support cartesian product of
>> 2
>>>>> data set(one of which can be infinite stream).
>>>>> For example, A(parallelism=2) broadcast join B(parallelism=2) in
>> JobVertex
>>>>> C.
>>>>> The idea to is have 4 C subtasks to deal with different combinations
>> of A/B
>>>>> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
>>>>> This requires one record to be sent to multiple downstream subtasks,
>> but
>>>>> not to all subtasks.
>>>>>
>>>>> With current interface this is not supported, as one record can only
>> be
>>>>> sent to one subtask, or to all subtasks of a JobVertex.
>>>>> And the user had to split the broadcast data set manually to several
>>>>> different JobVertices, which is hard to maintain and extend.
>>>>>
>>>>> Thanks,
>>>>> Zhu Zhu
>>>>>
>>>>> Yun Gao <[hidden email] <mailto:
>> [hidden email] <mailto:[hidden email]>>>
>> 于2019年8月22日周四 下午8:42写道：
>>>>>
>>>>> Hi everyone,
>>>>>    In some scenarios we met a requirement that some operators want to
>>>>> send records to theirs downstream operators with an multicast
>> communication
>>>>> pattern. In detail, for some records, the operators want to send them
>>>>> according to the partitioner (for example, Rebalance), and for some
>> other
>>>>> records, the operators want to send them to all the connected
>> operators and
>>>>> tasks. Such a communication pattern could be viewed as a kind of
>> multicast:
>>>>> it does not broadcast every record, but some record will indeed be
>> sent to
>>>>> multiple downstream operators.
>>>>>
>>>>> However, we found that this kind of communication pattern seems could
>> not
>>>>> be implemented rightly if the operators have multiple consumers with
>>>>> different parallelism, using the customized partitioner. To solve the
>> above
>>>>> problem, we propose to enhance the support for such kind of irregular
>>>>> communication pattern. We think there may be two options:
>>>>>
>>>>>   1. Support a kind of customized operator events, which share much
>>>>> similarity with Watermark, and these events can be broadcasted to the
>>>>> downstream operators separately.
>>>>>   2. Let the channel selector supports multicast, and also add the
>>>>> separate RecordWriter implementation to avoid impacting the
>> performance of
>>>>> the channel selector that does not need multicast.
>>>>>
>>>>> The problem and options are detailed in
>>>>>
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>> <
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing>
>> <
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>> <
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>>>>
>>>>>
>>>>> We are also wondering if there are other methods to implement this
>>>>> requirement with or without changing Runtime. Very thanks for any
>> feedbacks
>>>>> !
>>>>>
>>>>>
>>>>> Best,
>>>>> Yun
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>
>>
>>

[SHI Xiaogang]

Hi all,

I also think that multicasting is a necessity in Flink, but more details
are needed to be considered.

Currently network is tightly coupled with states in Flink to achieve
automatic scaling. We can only access keyed states in keyed streams and
operator states in all streams.
In the concrete example of theta-joins implemented with mutlticasting, the
following questions exist:

   - In which type of states will the data be stored? Do we need another
   type of states which is coupled with multicasting streams?
   - How to ensure the consistency between network and states when jobs
   scale out or scale in?

Regards,
Xiaogang

Xingcan Cui <[hidden email]> 于2019年8月25日周日 上午10:03写道：

> Hi all,
>
> Sorry for joining this thread late. Basically, I think enabling multicast
> pattern could be the right direction, but more detailed implementation
> policies need to be discussed.
>
> Two years ago, I filed an issue [1] about the multicast API. However, due
> to some reasons, it was laid aside. After that, when I tried to cherry-pick
> the change for experimental use, I found the return type of
> `selectChannels()` method had changed from `int[]` to `int`, which makes
> the old implementation not work anymore.
>
> From my side, the multicast has always been used for theta-join. As far as
> I know, it’s an essential requirement for some sophisticated joining
> algorithms. Until now, the Flink non-equi joins can still only be executed
> single-threaded. If we'd like to make some improvements on this, we should
> first take some measures to support multicast pattern.
>
> Best,
> Xingcan
>
> [1] https://issues.apache.org/jira/browse/FLINK-6936
>
> > On Aug 24, 2019, at 5:54 AM, Zhu Zhu <[hidden email]> wrote:
> >
> > Hi Piotr,
> >
> > Thanks for the explanation.
> > Agreed that the broadcastEmit(record) is a better choice for broadcasting
> > for the iterations.
> > As broadcasting for the iterations is the first motivation, let's support
> > it first.
> >
> > Thanks,
> > Zhu Zhu
> >
> > Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：
> >
> >>     Hi Piotr,
> >>
> >>      Very thanks for the suggestions!
> >>
> >>     Totally agree with that we could first focus on the broadcast
> >> scenarios and exposing the broadcastEmit method first considering the
> >> semantics and performance.
> >>
> >>     For the keyed stream, I also agree with that broadcasting keyed
> >> records to all the tasks may be confused considering the semantics of
> keyed
> >> partitioner. However, in the iteration case supporting broadcast over
> keyed
> >> partitioner should be required since users may create any subgraph for
> the
> >> iteration body, including the operators with key. I think a possible
> >> solution to this issue is to introduce another data type for
> >> 'broadcastEmit'. For example, for an operator Operator<T>, it may
> broadcast
> >> emit another type E instead of T, and the transmitting E will bypass the
> >> partitioner and setting keyed context. This should result in the design
> to
> >> introduce customized operator event (option 1 in the document). The
> cost of
> >> this method is that we need to introduce a new type of StreamElement and
> >> new interface for this type, but it should be suitable for both keyed or
> >> non-keyed partitioner.
> >>
> >> Best,
> >> Yun
> >>
> >>
> >>
> >> ------------------------------------------------------------------
> >> From:Piotr Nowojski <[hidden email]>
> >> Send Time:2019 Aug. 23 (Fri.) 22:29
> >> To:Zhu Zhu <[hidden email]>
> >> Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
> >> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> Pattern
> >>
> >> Hi,
> >>
> >> If the primary motivation is broadcasting (for the iterations) and we
> have
> >> no immediate need for multicast (cross join), I would prefer to first
> >> expose broadcast via the DataStream API and only later, once we finally
> >> need it, support multicast. As I wrote, multicast would be more
> challenging
> >> to implement, with more complicated runtime and API. And re-using
> multicast
> >> just to support broadcast doesn’t have much sense:
> >>
> >> 1. It’s a bit obfuscated. It’s easier to understand
> >> collectBroadcast(record) or broadcastEmit(record) compared to some
> >> multicast channel selector that just happens to return all of the
> channels.
> >> 2. There are performance benefits of explicitly calling
> >> `RecordWriter#broadcastEmit`.
> >>
> >>
> >> On a different note, what would be the semantic of such broadcast emit
> on
> >> KeyedStream? Would it be supported? Or would we limit support only to
> the
> >> non-keyed streams?
> >>
> >> Piotrek
> >>
> >>> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
> >>>
> >>> Thanks Piotr,
> >>>
> >>> Users asked for this feature sometimes ago when they migrating batch
> >> jobs to Flink(Blink).
> >>> It's not very urgent as they have taken some workarounds to solve
> >> it.(like partitioning data set to different job vertices)
> >>> So it's fine to not make it top priority.
> >>>
> >>> Anyway, as a commonly known scenario, I think users can benefit from
> >> cross join sooner or later.
> >>>
> >>> Thanks,
> >>> Zhu Zhu
> >>>
> >>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> >> 于2019年8月23日周五 下午6:19写道：
> >>> Hi,
> >>>
> >>> Thanks for the answers :) Ok I understand the full picture now. +1 from
> >> my side on solving this issue somehow. But before we start discussing
> how
> >> to solve it one last control question:
> >>>
> >>> I guess this multicast is intended to be used in blink planner, right?
> >> Assuming that we implement the multicast support now, when would it be
> used
> >> by the blink? I would like to avoid a scenario, where we implement an
> >> unused feature and we keep maintaining it for a long period of time.
> >>>
> >>> Piotrek
> >>>
> >>> PS, try to include motivating examples, including concrete ones in the
> >> proposals/design docs, for example in the very first paragraph.
> Especially
> >> if it’s a commonly known feature like cross join :)
> >>>
> >>>> On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
> >> wrote:
> >>>>
> >>>>    Hi Piotr,
> >>>>
> >>>>       Thanks a lot for sharing the thoughts!
> >>>>
> >>>>       For the iteration, agree with that multicasting is not
> >> necessary. Exploring the broadcast interface to Output of the operators
> in
> >> some way should also solve this issue, and I think it should be even
> more
> >> convenient to have the broadcast method for the iteration.
> >>>>
> >>>>       Also thanks Zhu Zhu for the cross join case!
> >>>> Best,
> >>>>  Yun
> >>>>
> >>>>
> >>>>
> >>>> ------------------------------------------------------------------
> >>>> From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
> >>>> Send Time:2019 Aug. 23 (Fri.) 17:25
> >>>> To:dev <[hidden email] <mailto:[hidden email]>>
> >>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> >>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> >> Pattern
> >>>>
> >>>> Hi Piotr,
> >>>>
> >>>> Yes you are right it's a distributed cross join requirement.
> >>>> Broadcast join can help with cross join cases. But users cannot use it
> >> if the data set to join is too large to fit into one subtask.
> >>>>
> >>>> Sorry for left some details behind.
> >>>>
> >>>> Thanks,
> >>>> Zhu Zhu
> >>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> >> 于2019年8月23日周五 下午4:57写道：
> >>>> Hi Yun and Zhu Zhu,
> >>>>
> >>>> Thanks for the more detailed example Zhu Zhu.
> >>>>
> >>>> As far as I understand for the iterations example we do not need
> >> multicasting. Regarding the Join example, I don’t fully understand it.
> The
> >> example that Zhu Zhu presented has a drawback of sending both tables to
> >> multiple nodes. What’s the benefit of using broadcast join over a hash
> join
> >> in such case? As far as I know, the biggest benefit of using broadcast
> join
> >> instead of hash join is that we can avoid sending the larger table over
> the
> >> network, because we can perform the join locally. In this example we are
> >> sending both of the tables to multiple nodes, which should defeat the
> >> purpose.
> >>>>
> >>>> Is it about implementing cross join or near cross joins in a
> >> distributed fashion?
> >>>>
> >>>>> if we introduce a new MulticastRecordWriter
> >>>>
> >>>> That’s one of the solutions. It might have a drawback of 3 class
> >> virtualisation problem (We have RecordWriter and BroadcastRecordWriter
> >> already). With up to two implementations, JVM is able to devirtualise
> the
> >> calls.
> >>>>
> >>>> Previously I was also thinking about just providing two different
> >> ChannelSelector interfaces. One with `int[]` and `SingleChannelSelector`
> >> with plain `int` and based on that, RecordWriter could perform some
> magic
> >> (worst case scenario `instaceof` checks).
> >>>>
> >>>> Another solution might be to change `ChannelSelector` interface into
> >> an iterator.
> >>>>
> >>>> But let's discuss the details after we agree on implementing this.
> >>>>
> >>>> Piotrek
> >>>>
> >>>>> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
> >> [hidden email]>> wrote:
> >>>>>
> >>>>>  Hi Piotr,
> >>>>>
> >>>>>       Thanks a lot for the suggestions!
> >>>>>
> >>>>>       The core motivation of this discussion is to implement a new
> >> iteration library on the DataStream, and it requires to insert special
> >> records in the stream to notify the progress of the iteration. The
> >> mechanism of such records is very similar to the current Watermark, and
> we
> >> meet the problem of sending normal records according to the partition
> >> (Rebalance, etc..) and also be able to broadcast the inserted progress
> >> records to all the connected records. I have read the notes in the
> google
> >> doc and I totally agree with that exploring the broadcast interface in
> >> RecordWriter in some way is able to solve this issue.
> >>>>>
> >>>>>      Regarding to `int[] ChannelSelector#selectChannels()`, I'm
> >> wondering if we introduce a new MulticastRecordWriter and left the
> current
> >> RecordWriter untouched, could we avoid the performance degradation ?
> Since
> >> with such a modification the normal RecordWriter does not need to
> iterate
> >> the return array by ChannelSelector, and the only difference will be
> >> returning an array instead of an integer, and accessing the first
> element
> >> of the returned array instead of reading the integer directly.
> >>>>>
> >>>>> Best,
> >>>>> Yun
> >>>>>
> >>>>> ------------------------------------------------------------------
> >>>>> From:Piotr Nowojski <[hidden email] <mailto:[hidden email]
> >>>>
> >>>>> Send Time:2019 Aug. 23 (Fri.) 15:20
> >>>>> To:dev <[hidden email] <mailto:[hidden email]>>
> >>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> >>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> >> Pattern
> >>>>>
> >>>>> Hi,
> >>>>>
> >>>>> Yun:
> >>>>>
> >>>>> Thanks for proposing the idea. I have checked the document and left
> >> couple of questions there, but it might be better to answer them here.
> >>>>>
> >>>>> What is the exact motivation and what problems do you want to solve?
> >> We have dropped multicast support from the network stack [1] for two
> >> reasons:
> >>>>> 1. Performance
> >>>>> 2. Code simplicity
> >>>>>
> >>>>> The proposal to re introduce `int[] ChannelSelector#selectChannels()`
> >> would revert those changes. At that time we were thinking about a way
> how
> >> to keep the multicast support on the network level, while keeping the
> >> performance and simplicity for non multicast cases and there are ways to
> >> achieve that. However they would add extra complexity to Flink, which it
> >> would be better to avoid.
> >>>>>
> >>>>> On the other hand, supporting dual pattern: standard partitioning or
> >> broadcasting is easy to do, as LatencyMarkers are doing exactly that. It
> >> would be just a matter of exposing this to the user in some way. So
> before
> >> we go any further, can you describe your use cases/motivation? Isn’t
> mix of
> >> standard partitioning and broadcasting enough? Do we need multicasting?
> >>>>>
> >>>>> Zhu:
> >>>>>
> >>>>> Could you rephrase your example? I didn’t quite understand it.
> >>>>>
> >>>>> Piotrek
> >>>>>
> >>>>> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> >> https://issues.apache.org/jira/browse/FLINK-10662> <
> >> https://issues.apache.org/jira/browse/FLINK-10662 <
> >> https://issues.apache.org/jira/browse/FLINK-10662>>
> >>>>>
> >>>>> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
> >> [hidden email]> <mailto:[hidden email] <mailto:[hidden email]
> >>>
> >> wrote:
> >>>>>
> >>>>> Thanks Yun for starting this discussion.
> >>>>> I think the multicasting can be very helpful in certain cases.
> >>>>>
> >>>>> I have received requirements from users that they want to do
> broadcast
> >>>>> join, while the data set to broadcast is too large to fit in one
> task.
> >>>>> Thus the requirement turned out to be to support cartesian product of
> >> 2
> >>>>> data set(one of which can be infinite stream).
> >>>>> For example, A(parallelism=2) broadcast join B(parallelism=2) in
> >> JobVertex
> >>>>> C.
> >>>>> The idea to is have 4 C subtasks to deal with different combinations
> >> of A/B
> >>>>> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> >>>>> This requires one record to be sent to multiple downstream subtasks,
> >> but
> >>>>> not to all subtasks.
> >>>>>
> >>>>> With current interface this is not supported, as one record can only
> >> be
> >>>>> sent to one subtask, or to all subtasks of a JobVertex.
> >>>>> And the user had to split the broadcast data set manually to several
> >>>>> different JobVertices, which is hard to maintain and extend.
> >>>>>
> >>>>> Thanks,
> >>>>> Zhu Zhu
> >>>>>
> >>>>> Yun Gao <[hidden email] <mailto:
> >> [hidden email] <mailto:[hidden email]>>>
> >> 于2019年8月22日周四 下午8:42写道：
> >>>>>
> >>>>> Hi everyone,
> >>>>>    In some scenarios we met a requirement that some operators want to
> >>>>> send records to theirs downstream operators with an multicast
> >> communication
> >>>>> pattern. In detail, for some records, the operators want to send them
> >>>>> according to the partitioner (for example, Rebalance), and for some
> >> other
> >>>>> records, the operators want to send them to all the connected
> >> operators and
> >>>>> tasks. Such a communication pattern could be viewed as a kind of
> >> multicast:
> >>>>> it does not broadcast every record, but some record will indeed be
> >> sent to
> >>>>> multiple downstream operators.
> >>>>>
> >>>>> However, we found that this kind of communication pattern seems could
> >> not
> >>>>> be implemented rightly if the operators have multiple consumers with
> >>>>> different parallelism, using the customized partitioner. To solve the
> >> above
> >>>>> problem, we propose to enhance the support for such kind of irregular
> >>>>> communication pattern. We think there may be two options:
> >>>>>
> >>>>>   1. Support a kind of customized operator events, which share much
> >>>>> similarity with Watermark, and these events can be broadcasted to the
> >>>>> downstream operators separately.
> >>>>>   2. Let the channel selector supports multicast, and also add the
> >>>>> separate RecordWriter implementation to avoid impacting the
> >> performance of
> >>>>> the channel selector that does not need multicast.
> >>>>>
> >>>>> The problem and options are detailed in
> >>>>>
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >> <
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >
> >> <
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >> <
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >>>>
> >>>>>
> >>>>> We are also wondering if there are other methods to implement this
> >>>>> requirement with or without changing Runtime. Very thanks for any
> >> feedbacks
> >>>>> !
> >>>>>
> >>>>>
> >>>>> Best,
> >>>>> Yun
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>
> >>>>
> >>>
> >>
> >>
> >>
>
>

[Guowei Ma]

Thanks Yun for bringing up this discussion and very thanks for all the deep
thoughts!

For now, I think this discussion contains two scenarios: one if for
iteration library support and the other is for SQL join support. I think
both of the two scenarios are useful but they seem to have different best
suitable solutions. For making the discussion more clear, I would suggest
to split the discussion into two threads.

And I agree with Piotr that it is very tricky that a keyed stream received
a "broadcast element". So we may add some new interfaces, which could
broadcast or process some special "broadcast event". In that way "broadcast
event" will not be sent with the normal process.

Best,
Guowei


SHI Xiaogang <[hidden email]> 于2019年8月26日周一 上午9:27写道：

> Hi all,
>
> I also think that multicasting is a necessity in Flink, but more details
> are needed to be considered.
>
> Currently network is tightly coupled with states in Flink to achieve
> automatic scaling. We can only access keyed states in keyed streams and
> operator states in all streams.
> In the concrete example of theta-joins implemented with mutlticasting, the
> following questions exist:
>
>    - In which type of states will the data be stored? Do we need another
>    type of states which is coupled with multicasting streams?
>    - How to ensure the consistency between network and states when jobs
>    scale out or scale in?
>
> Regards,
> Xiaogang
>
> Xingcan Cui <[hidden email]> 于2019年8月25日周日 上午10:03写道：
>
> > Hi all,
> >
> > Sorry for joining this thread late. Basically, I think enabling multicast
> > pattern could be the right direction, but more detailed implementation
> > policies need to be discussed.
> >
> > Two years ago, I filed an issue [1] about the multicast API. However, due
> > to some reasons, it was laid aside. After that, when I tried to
> cherry-pick
> > the change for experimental use, I found the return type of
> > `selectChannels()` method had changed from `int[]` to `int`, which makes
> > the old implementation not work anymore.
> >
> > From my side, the multicast has always been used for theta-join. As far
> as
> > I know, it’s an essential requirement for some sophisticated joining
> > algorithms. Until now, the Flink non-equi joins can still only be
> executed
> > single-threaded. If we'd like to make some improvements on this, we
> should
> > first take some measures to support multicast pattern.
> >
> > Best,
> > Xingcan
> >
> > [1] https://issues.apache.org/jira/browse/FLINK-6936
> >
> > > On Aug 24, 2019, at 5:54 AM, Zhu Zhu <[hidden email]> wrote:
> > >
> > > Hi Piotr,
> > >
> > > Thanks for the explanation.
> > > Agreed that the broadcastEmit(record) is a better choice for
> broadcasting
> > > for the iterations.
> > > As broadcasting for the iterations is the first motivation, let's
> support
> > > it first.
> > >
> > > Thanks,
> > > Zhu Zhu
> > >
> > > Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：
> > >
> > >>     Hi Piotr,
> > >>
> > >>      Very thanks for the suggestions!
> > >>
> > >>     Totally agree with that we could first focus on the broadcast
> > >> scenarios and exposing the broadcastEmit method first considering the
> > >> semantics and performance.
> > >>
> > >>     For the keyed stream, I also agree with that broadcasting keyed
> > >> records to all the tasks may be confused considering the semantics of
> > keyed
> > >> partitioner. However, in the iteration case supporting broadcast over
> > keyed
> > >> partitioner should be required since users may create any subgraph for
> > the
> > >> iteration body, including the operators with key. I think a possible
> > >> solution to this issue is to introduce another data type for
> > >> 'broadcastEmit'. For example, for an operator Operator<T>, it may
> > broadcast
> > >> emit another type E instead of T, and the transmitting E will bypass
> the
> > >> partitioner and setting keyed context. This should result in the
> design
> > to
> > >> introduce customized operator event (option 1 in the document). The
> > cost of
> > >> this method is that we need to introduce a new type of StreamElement
> and
> > >> new interface for this type, but it should be suitable for both keyed
> or
> > >> non-keyed partitioner.
> > >>
> > >> Best,
> > >> Yun
> > >>
> > >>
> > >>
> > >> ------------------------------------------------------------------
> > >> From:Piotr Nowojski <[hidden email]>
> > >> Send Time:2019 Aug. 23 (Fri.) 22:29
> > >> To:Zhu Zhu <[hidden email]>
> > >> Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
> > >> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> > Pattern
> > >>
> > >> Hi,
> > >>
> > >> If the primary motivation is broadcasting (for the iterations) and we
> > have
> > >> no immediate need for multicast (cross join), I would prefer to first
> > >> expose broadcast via the DataStream API and only later, once we
> finally
> > >> need it, support multicast. As I wrote, multicast would be more
> > challenging
> > >> to implement, with more complicated runtime and API. And re-using
> > multicast
> > >> just to support broadcast doesn’t have much sense:
> > >>
> > >> 1. It’s a bit obfuscated. It’s easier to understand
> > >> collectBroadcast(record) or broadcastEmit(record) compared to some
> > >> multicast channel selector that just happens to return all of the
> > channels.
> > >> 2. There are performance benefits of explicitly calling
> > >> `RecordWriter#broadcastEmit`.
> > >>
> > >>
> > >> On a different note, what would be the semantic of such broadcast emit
> > on
> > >> KeyedStream? Would it be supported? Or would we limit support only to
> > the
> > >> non-keyed streams?
> > >>
> > >> Piotrek
> > >>
> > >>> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
> > >>>
> > >>> Thanks Piotr,
> > >>>
> > >>> Users asked for this feature sometimes ago when they migrating batch
> > >> jobs to Flink(Blink).
> > >>> It's not very urgent as they have taken some workarounds to solve
> > >> it.(like partitioning data set to different job vertices)
> > >>> So it's fine to not make it top priority.
> > >>>
> > >>> Anyway, as a commonly known scenario, I think users can benefit from
> > >> cross join sooner or later.
> > >>>
> > >>> Thanks,
> > >>> Zhu Zhu
> > >>>
> > >>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> > >> 于2019年8月23日周五 下午6:19写道：
> > >>> Hi,
> > >>>
> > >>> Thanks for the answers :) Ok I understand the full picture now. +1
> from
> > >> my side on solving this issue somehow. But before we start discussing
> > how
> > >> to solve it one last control question:
> > >>>
> > >>> I guess this multicast is intended to be used in blink planner,
> right?
> > >> Assuming that we implement the multicast support now, when would it be
> > used
> > >> by the blink? I would like to avoid a scenario, where we implement an
> > >> unused feature and we keep maintaining it for a long period of time.
> > >>>
> > >>> Piotrek
> > >>>
> > >>> PS, try to include motivating examples, including concrete ones in
> the
> > >> proposals/design docs, for example in the very first paragraph.
> > Especially
> > >> if it’s a commonly known feature like cross join :)
> > >>>
> > >>>> On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
> > >> wrote:
> > >>>>
> > >>>>    Hi Piotr,
> > >>>>
> > >>>>       Thanks a lot for sharing the thoughts!
> > >>>>
> > >>>>       For the iteration, agree with that multicasting is not
> > >> necessary. Exploring the broadcast interface to Output of the
> operators
> > in
> > >> some way should also solve this issue, and I think it should be even
> > more
> > >> convenient to have the broadcast method for the iteration.
> > >>>>
> > >>>>       Also thanks Zhu Zhu for the cross join case!
> > >>>> Best,
> > >>>>  Yun
> > >>>>
> > >>>>
> > >>>>
> > >>>> ------------------------------------------------------------------
> > >>>> From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
> > >>>> Send Time:2019 Aug. 23 (Fri.) 17:25
> > >>>> To:dev <[hidden email] <mailto:[hidden email]>>
> > >>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > >>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> > >> Pattern
> > >>>>
> > >>>> Hi Piotr,
> > >>>>
> > >>>> Yes you are right it's a distributed cross join requirement.
> > >>>> Broadcast join can help with cross join cases. But users cannot use
> it
> > >> if the data set to join is too large to fit into one subtask.
> > >>>>
> > >>>> Sorry for left some details behind.
> > >>>>
> > >>>> Thanks,
> > >>>> Zhu Zhu
> > >>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> > >> 于2019年8月23日周五 下午4:57写道：
> > >>>> Hi Yun and Zhu Zhu,
> > >>>>
> > >>>> Thanks for the more detailed example Zhu Zhu.
> > >>>>
> > >>>> As far as I understand for the iterations example we do not need
> > >> multicasting. Regarding the Join example, I don’t fully understand it.
> > The
> > >> example that Zhu Zhu presented has a drawback of sending both tables
> to
> > >> multiple nodes. What’s the benefit of using broadcast join over a hash
> > join
> > >> in such case? As far as I know, the biggest benefit of using broadcast
> > join
> > >> instead of hash join is that we can avoid sending the larger table
> over
> > the
> > >> network, because we can perform the join locally. In this example we
> are
> > >> sending both of the tables to multiple nodes, which should defeat the
> > >> purpose.
> > >>>>
> > >>>> Is it about implementing cross join or near cross joins in a
> > >> distributed fashion?
> > >>>>
> > >>>>> if we introduce a new MulticastRecordWriter
> > >>>>
> > >>>> That’s one of the solutions. It might have a drawback of 3 class
> > >> virtualisation problem (We have RecordWriter and BroadcastRecordWriter
> > >> already). With up to two implementations, JVM is able to devirtualise
> > the
> > >> calls.
> > >>>>
> > >>>> Previously I was also thinking about just providing two different
> > >> ChannelSelector interfaces. One with `int[]` and
> `SingleChannelSelector`
> > >> with plain `int` and based on that, RecordWriter could perform some
> > magic
> > >> (worst case scenario `instaceof` checks).
> > >>>>
> > >>>> Another solution might be to change `ChannelSelector` interface into
> > >> an iterator.
> > >>>>
> > >>>> But let's discuss the details after we agree on implementing this.
> > >>>>
> > >>>> Piotrek
> > >>>>
> > >>>>> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
> > >> [hidden email]>> wrote:
> > >>>>>
> > >>>>>  Hi Piotr,
> > >>>>>
> > >>>>>       Thanks a lot for the suggestions!
> > >>>>>
> > >>>>>       The core motivation of this discussion is to implement a new
> > >> iteration library on the DataStream, and it requires to insert special
> > >> records in the stream to notify the progress of the iteration. The
> > >> mechanism of such records is very similar to the current Watermark,
> and
> > we
> > >> meet the problem of sending normal records according to the partition
> > >> (Rebalance, etc..) and also be able to broadcast the inserted progress
> > >> records to all the connected records. I have read the notes in the
> > google
> > >> doc and I totally agree with that exploring the broadcast interface in
> > >> RecordWriter in some way is able to solve this issue.
> > >>>>>
> > >>>>>      Regarding to `int[] ChannelSelector#selectChannels()`, I'm
> > >> wondering if we introduce a new MulticastRecordWriter and left the
> > current
> > >> RecordWriter untouched, could we avoid the performance degradation ?
> > Since
> > >> with such a modification the normal RecordWriter does not need to
> > iterate
> > >> the return array by ChannelSelector, and the only difference will be
> > >> returning an array instead of an integer, and accessing the first
> > element
> > >> of the returned array instead of reading the integer directly.
> > >>>>>
> > >>>>> Best,
> > >>>>> Yun
> > >>>>>
> > >>>>> ------------------------------------------------------------------
> > >>>>> From:Piotr Nowojski <[hidden email] <mailto:
> [hidden email]
> > >>>>
> > >>>>> Send Time:2019 Aug. 23 (Fri.) 15:20
> > >>>>> To:dev <[hidden email] <mailto:[hidden email]>>
> > >>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > >>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> > >> Pattern
> > >>>>>
> > >>>>> Hi,
> > >>>>>
> > >>>>> Yun:
> > >>>>>
> > >>>>> Thanks for proposing the idea. I have checked the document and left
> > >> couple of questions there, but it might be better to answer them here.
> > >>>>>
> > >>>>> What is the exact motivation and what problems do you want to
> solve?
> > >> We have dropped multicast support from the network stack [1] for two
> > >> reasons:
> > >>>>> 1. Performance
> > >>>>> 2. Code simplicity
> > >>>>>
> > >>>>> The proposal to re introduce `int[]
> ChannelSelector#selectChannels()`
> > >> would revert those changes. At that time we were thinking about a way
> > how
> > >> to keep the multicast support on the network level, while keeping the
> > >> performance and simplicity for non multicast cases and there are ways
> to
> > >> achieve that. However they would add extra complexity to Flink, which
> it
> > >> would be better to avoid.
> > >>>>>
> > >>>>> On the other hand, supporting dual pattern: standard partitioning
> or
> > >> broadcasting is easy to do, as LatencyMarkers are doing exactly that.
> It
> > >> would be just a matter of exposing this to the user in some way. So
> > before
> > >> we go any further, can you describe your use cases/motivation? Isn’t
> > mix of
> > >> standard partitioning and broadcasting enough? Do we need
> multicasting?
> > >>>>>
> > >>>>> Zhu:
> > >>>>>
> > >>>>> Could you rephrase your example? I didn’t quite understand it.
> > >>>>>
> > >>>>> Piotrek
> > >>>>>
> > >>>>> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> > >> https://issues.apache.org/jira/browse/FLINK-10662> <
> > >> https://issues.apache.org/jira/browse/FLINK-10662 <
> > >> https://issues.apache.org/jira/browse/FLINK-10662>>
> > >>>>>
> > >>>>> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
> > >> [hidden email]> <mailto:[hidden email] <mailto:
> [hidden email]
> > >>>
> > >> wrote:
> > >>>>>
> > >>>>> Thanks Yun for starting this discussion.
> > >>>>> I think the multicasting can be very helpful in certain cases.
> > >>>>>
> > >>>>> I have received requirements from users that they want to do
> > broadcast
> > >>>>> join, while the data set to broadcast is too large to fit in one
> > task.
> > >>>>> Thus the requirement turned out to be to support cartesian product
> of
> > >> 2
> > >>>>> data set(one of which can be infinite stream).
> > >>>>> For example, A(parallelism=2) broadcast join B(parallelism=2) in
> > >> JobVertex
> > >>>>> C.
> > >>>>> The idea to is have 4 C subtasks to deal with different
> combinations
> > >> of A/B
> > >>>>> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> > >>>>> This requires one record to be sent to multiple downstream
> subtasks,
> > >> but
> > >>>>> not to all subtasks.
> > >>>>>
> > >>>>> With current interface this is not supported, as one record can
> only
> > >> be
> > >>>>> sent to one subtask, or to all subtasks of a JobVertex.
> > >>>>> And the user had to split the broadcast data set manually to
> several
> > >>>>> different JobVertices, which is hard to maintain and extend.
> > >>>>>
> > >>>>> Thanks,
> > >>>>> Zhu Zhu
> > >>>>>
> > >>>>> Yun Gao <[hidden email] <mailto:
> > >> [hidden email] <mailto:[hidden email]>>>
> > >> 于2019年8月22日周四 下午8:42写道：
> > >>>>>
> > >>>>> Hi everyone,
> > >>>>>    In some scenarios we met a requirement that some operators want
> to
> > >>>>> send records to theirs downstream operators with an multicast
> > >> communication
> > >>>>> pattern. In detail, for some records, the operators want to send
> them
> > >>>>> according to the partitioner (for example, Rebalance), and for some
> > >> other
> > >>>>> records, the operators want to send them to all the connected
> > >> operators and
> > >>>>> tasks. Such a communication pattern could be viewed as a kind of
> > >> multicast:
> > >>>>> it does not broadcast every record, but some record will indeed be
> > >> sent to
> > >>>>> multiple downstream operators.
> > >>>>>
> > >>>>> However, we found that this kind of communication pattern seems
> could
> > >> not
> > >>>>> be implemented rightly if the operators have multiple consumers
> with
> > >>>>> different parallelism, using the customized partitioner. To solve
> the
> > >> above
> > >>>>> problem, we propose to enhance the support for such kind of
> irregular
> > >>>>> communication pattern. We think there may be two options:
> > >>>>>
> > >>>>>   1. Support a kind of customized operator events, which share much
> > >>>>> similarity with Watermark, and these events can be broadcasted to
> the
> > >>>>> downstream operators separately.
> > >>>>>   2. Let the channel selector supports multicast, and also add the
> > >>>>> separate RecordWriter implementation to avoid impacting the
> > >> performance of
> > >>>>> the channel selector that does not need multicast.
> > >>>>>
> > >>>>> The problem and options are detailed in
> > >>>>>
> > >>
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > >> <
> > >>
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > >
> > >> <
> > >>
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > >> <
> > >>
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > >>>>
> > >>>>>
> > >>>>> We are also wondering if there are other methods to implement this
> > >>>>> requirement with or without changing Runtime. Very thanks for any
> > >> feedbacks
> > >>>>> !
> > >>>>>
> > >>>>>
> > >>>>> Best,
> > >>>>> Yun
> > >>>>>
> > >>>>>
> > >>>>>
> > >>>>>
> > >>>>
> > >>>>
> > >>>
> > >>
> > >>
> > >>
> >
> >
>

[Kurt Young]

From SQL's perspective, distributed cross join is a valid feature but not
very
urgent. Actually this discuss reminds me about another useful feature
(sorry
for the distraction):

when doing broadcast in batch shuffle mode, we can make each producer only
write one copy of the output data, but not for every consumer. Broadcast
join
is much more useful, and this is a very important optimization. Not sure if
we
have already consider this.

Best,
Kurt


On Mon, Aug 26, 2019 at 12:16 PM Guowei Ma <[hidden email]> wrote:

> Thanks Yun for bringing up this discussion and very thanks for all the deep
> thoughts!
>
> For now, I think this discussion contains two scenarios: one if for
> iteration library support and the other is for SQL join support. I think
> both of the two scenarios are useful but they seem to have different best
> suitable solutions. For making the discussion more clear, I would suggest
> to split the discussion into two threads.
>
> And I agree with Piotr that it is very tricky that a keyed stream received
> a "broadcast element". So we may add some new interfaces, which could
> broadcast or process some special "broadcast event". In that way "broadcast
> event" will not be sent with the normal process.
>
> Best,
> Guowei
>
>
> SHI Xiaogang <[hidden email]> 于2019年8月26日周一 上午9:27写道：
>
> > Hi all,
> >
> > I also think that multicasting is a necessity in Flink, but more details
> > are needed to be considered.
> >
> > Currently network is tightly coupled with states in Flink to achieve
> > automatic scaling. We can only access keyed states in keyed streams and
> > operator states in all streams.
> > In the concrete example of theta-joins implemented with mutlticasting,
> the
> > following questions exist:
> >
> >    - In which type of states will the data be stored? Do we need another
> >    type of states which is coupled with multicasting streams?
> >    - How to ensure the consistency between network and states when jobs
> >    scale out or scale in?
> >
> > Regards,
> > Xiaogang
> >
> > Xingcan Cui <[hidden email]> 于2019年8月25日周日 上午10:03写道：
> >
> > > Hi all,
> > >
> > > Sorry for joining this thread late. Basically, I think enabling
> multicast
> > > pattern could be the right direction, but more detailed implementation
> > > policies need to be discussed.
> > >
> > > Two years ago, I filed an issue [1] about the multicast API. However,
> due
> > > to some reasons, it was laid aside. After that, when I tried to
> > cherry-pick
> > > the change for experimental use, I found the return type of
> > > `selectChannels()` method had changed from `int[]` to `int`, which
> makes
> > > the old implementation not work anymore.
> > >
> > > From my side, the multicast has always been used for theta-join. As far
> > as
> > > I know, it’s an essential requirement for some sophisticated joining
> > > algorithms. Until now, the Flink non-equi joins can still only be
> > executed
> > > single-threaded. If we'd like to make some improvements on this, we
> > should
> > > first take some measures to support multicast pattern.
> > >
> > > Best,
> > > Xingcan
> > >
> > > [1] https://issues.apache.org/jira/browse/FLINK-6936
> > >
> > > > On Aug 24, 2019, at 5:54 AM, Zhu Zhu <[hidden email]> wrote:
> > > >
> > > > Hi Piotr,
> > > >
> > > > Thanks for the explanation.
> > > > Agreed that the broadcastEmit(record) is a better choice for
> > broadcasting
> > > > for the iterations.
> > > > As broadcasting for the iterations is the first motivation, let's
> > support
> > > > it first.
> > > >
> > > > Thanks,
> > > > Zhu Zhu
> > > >
> > > > Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：
> > > >
> > > >>     Hi Piotr,
> > > >>
> > > >>      Very thanks for the suggestions!
> > > >>
> > > >>     Totally agree with that we could first focus on the broadcast
> > > >> scenarios and exposing the broadcastEmit method first considering
> the
> > > >> semantics and performance.
> > > >>
> > > >>     For the keyed stream, I also agree with that broadcasting keyed
> > > >> records to all the tasks may be confused considering the semantics
> of
> > > keyed
> > > >> partitioner. However, in the iteration case supporting broadcast
> over
> > > keyed
> > > >> partitioner should be required since users may create any subgraph
> for
> > > the
> > > >> iteration body, including the operators with key. I think a possible
> > > >> solution to this issue is to introduce another data type for
> > > >> 'broadcastEmit'. For example, for an operator Operator<T>, it may
> > > broadcast
> > > >> emit another type E instead of T, and the transmitting E will bypass
> > the
> > > >> partitioner and setting keyed context. This should result in the
> > design
> > > to
> > > >> introduce customized operator event (option 1 in the document). The
> > > cost of
> > > >> this method is that we need to introduce a new type of StreamElement
> > and
> > > >> new interface for this type, but it should be suitable for both
> keyed
> > or
> > > >> non-keyed partitioner.
> > > >>
> > > >> Best,
> > > >> Yun
> > > >>
> > > >>
> > > >>
> > > >> ------------------------------------------------------------------
> > > >> From:Piotr Nowojski <[hidden email]>
> > > >> Send Time:2019 Aug. 23 (Fri.) 22:29
> > > >> To:Zhu Zhu <[hidden email]>
> > > >> Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
> > > >> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> > > Pattern
> > > >>
> > > >> Hi,
> > > >>
> > > >> If the primary motivation is broadcasting (for the iterations) and
> we
> > > have
> > > >> no immediate need for multicast (cross join), I would prefer to
> first
> > > >> expose broadcast via the DataStream API and only later, once we
> > finally
> > > >> need it, support multicast. As I wrote, multicast would be more
> > > challenging
> > > >> to implement, with more complicated runtime and API. And re-using
> > > multicast
> > > >> just to support broadcast doesn’t have much sense:
> > > >>
> > > >> 1. It’s a bit obfuscated. It’s easier to understand
> > > >> collectBroadcast(record) or broadcastEmit(record) compared to some
> > > >> multicast channel selector that just happens to return all of the
> > > channels.
> > > >> 2. There are performance benefits of explicitly calling
> > > >> `RecordWriter#broadcastEmit`.
> > > >>
> > > >>
> > > >> On a different note, what would be the semantic of such broadcast
> emit
> > > on
> > > >> KeyedStream? Would it be supported? Or would we limit support only
> to
> > > the
> > > >> non-keyed streams?
> > > >>
> > > >> Piotrek
> > > >>
> > > >>> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
> > > >>>
> > > >>> Thanks Piotr,
> > > >>>
> > > >>> Users asked for this feature sometimes ago when they migrating
> batch
> > > >> jobs to Flink(Blink).
> > > >>> It's not very urgent as they have taken some workarounds to solve
> > > >> it.(like partitioning data set to different job vertices)
> > > >>> So it's fine to not make it top priority.
> > > >>>
> > > >>> Anyway, as a commonly known scenario, I think users can benefit
> from
> > > >> cross join sooner or later.
> > > >>>
> > > >>> Thanks,
> > > >>> Zhu Zhu
> > > >>>
> > > >>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> > > >> 于2019年8月23日周五 下午6:19写道：
> > > >>> Hi,
> > > >>>
> > > >>> Thanks for the answers :) Ok I understand the full picture now. +1
> > from
> > > >> my side on solving this issue somehow. But before we start
> discussing
> > > how
> > > >> to solve it one last control question:
> > > >>>
> > > >>> I guess this multicast is intended to be used in blink planner,
> > right?
> > > >> Assuming that we implement the multicast support now, when would it
> be
> > > used
> > > >> by the blink? I would like to avoid a scenario, where we implement
> an
> > > >> unused feature and we keep maintaining it for a long period of time.
> > > >>>
> > > >>> Piotrek
> > > >>>
> > > >>> PS, try to include motivating examples, including concrete ones in
> > the
> > > >> proposals/design docs, for example in the very first paragraph.
> > > Especially
> > > >> if it’s a commonly known feature like cross join :)
> > > >>>
> > > >>>> On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
> > > >> wrote:
> > > >>>>
> > > >>>>    Hi Piotr,
> > > >>>>
> > > >>>>       Thanks a lot for sharing the thoughts!
> > > >>>>
> > > >>>>       For the iteration, agree with that multicasting is not
> > > >> necessary. Exploring the broadcast interface to Output of the
> > operators
> > > in
> > > >> some way should also solve this issue, and I think it should be even
> > > more
> > > >> convenient to have the broadcast method for the iteration.
> > > >>>>
> > > >>>>       Also thanks Zhu Zhu for the cross join case!
> > > >>>> Best,
> > > >>>>  Yun
> > > >>>>
> > > >>>>
> > > >>>>
> > > >>>> ------------------------------------------------------------------
> > > >>>> From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
> > > >>>> Send Time:2019 Aug. 23 (Fri.) 17:25
> > > >>>> To:dev <[hidden email] <mailto:[hidden email]>>
> > > >>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > > >>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> > > >> Pattern
> > > >>>>
> > > >>>> Hi Piotr,
> > > >>>>
> > > >>>> Yes you are right it's a distributed cross join requirement.
> > > >>>> Broadcast join can help with cross join cases. But users cannot
> use
> > it
> > > >> if the data set to join is too large to fit into one subtask.
> > > >>>>
> > > >>>> Sorry for left some details behind.
> > > >>>>
> > > >>>> Thanks,
> > > >>>> Zhu Zhu
> > > >>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> > > >> 于2019年8月23日周五 下午4:57写道：
> > > >>>> Hi Yun and Zhu Zhu,
> > > >>>>
> > > >>>> Thanks for the more detailed example Zhu Zhu.
> > > >>>>
> > > >>>> As far as I understand for the iterations example we do not need
> > > >> multicasting. Regarding the Join example, I don’t fully understand
> it.
> > > The
> > > >> example that Zhu Zhu presented has a drawback of sending both tables
> > to
> > > >> multiple nodes. What’s the benefit of using broadcast join over a
> hash
> > > join
> > > >> in such case? As far as I know, the biggest benefit of using
> broadcast
> > > join
> > > >> instead of hash join is that we can avoid sending the larger table
> > over
> > > the
> > > >> network, because we can perform the join locally. In this example we
> > are
> > > >> sending both of the tables to multiple nodes, which should defeat
> the
> > > >> purpose.
> > > >>>>
> > > >>>> Is it about implementing cross join or near cross joins in a
> > > >> distributed fashion?
> > > >>>>
> > > >>>>> if we introduce a new MulticastRecordWriter
> > > >>>>
> > > >>>> That’s one of the solutions. It might have a drawback of 3 class
> > > >> virtualisation problem (We have RecordWriter and
> BroadcastRecordWriter
> > > >> already). With up to two implementations, JVM is able to
> devirtualise
> > > the
> > > >> calls.
> > > >>>>
> > > >>>> Previously I was also thinking about just providing two different
> > > >> ChannelSelector interfaces. One with `int[]` and
> > `SingleChannelSelector`
> > > >> with plain `int` and based on that, RecordWriter could perform some
> > > magic
> > > >> (worst case scenario `instaceof` checks).
> > > >>>>
> > > >>>> Another solution might be to change `ChannelSelector` interface
> into
> > > >> an iterator.
> > > >>>>
> > > >>>> But let's discuss the details after we agree on implementing this.
> > > >>>>
> > > >>>> Piotrek
> > > >>>>
> > > >>>>> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
> > > >> [hidden email]>> wrote:
> > > >>>>>
> > > >>>>>  Hi Piotr,
> > > >>>>>
> > > >>>>>       Thanks a lot for the suggestions!
> > > >>>>>
> > > >>>>>       The core motivation of this discussion is to implement a
> new
> > > >> iteration library on the DataStream, and it requires to insert
> special
> > > >> records in the stream to notify the progress of the iteration. The
> > > >> mechanism of such records is very similar to the current Watermark,
> > and
> > > we
> > > >> meet the problem of sending normal records according to the
> partition
> > > >> (Rebalance, etc..) and also be able to broadcast the inserted
> progress
> > > >> records to all the connected records. I have read the notes in the
> > > google
> > > >> doc and I totally agree with that exploring the broadcast interface
> in
> > > >> RecordWriter in some way is able to solve this issue.
> > > >>>>>
> > > >>>>>      Regarding to `int[] ChannelSelector#selectChannels()`, I'm
> > > >> wondering if we introduce a new MulticastRecordWriter and left the
> > > current
> > > >> RecordWriter untouched, could we avoid the performance degradation ?
> > > Since
> > > >> with such a modification the normal RecordWriter does not need to
> > > iterate
> > > >> the return array by ChannelSelector, and the only difference will be
> > > >> returning an array instead of an integer, and accessing the first
> > > element
> > > >> of the returned array instead of reading the integer directly.
> > > >>>>>
> > > >>>>> Best,
> > > >>>>> Yun
> > > >>>>>
> > > >>>>>
> ------------------------------------------------------------------
> > > >>>>> From:Piotr Nowojski <[hidden email] <mailto:
> > [hidden email]
> > > >>>>
> > > >>>>> Send Time:2019 Aug. 23 (Fri.) 15:20
> > > >>>>> To:dev <[hidden email] <mailto:[hidden email]>>
> > > >>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> > > >>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> > > >> Pattern
> > > >>>>>
> > > >>>>> Hi,
> > > >>>>>
> > > >>>>> Yun:
> > > >>>>>
> > > >>>>> Thanks for proposing the idea. I have checked the document and
> left
> > > >> couple of questions there, but it might be better to answer them
> here.
> > > >>>>>
> > > >>>>> What is the exact motivation and what problems do you want to
> > solve?
> > > >> We have dropped multicast support from the network stack [1] for two
> > > >> reasons:
> > > >>>>> 1. Performance
> > > >>>>> 2. Code simplicity
> > > >>>>>
> > > >>>>> The proposal to re introduce `int[]
> > ChannelSelector#selectChannels()`
> > > >> would revert those changes. At that time we were thinking about a
> way
> > > how
> > > >> to keep the multicast support on the network level, while keeping
> the
> > > >> performance and simplicity for non multicast cases and there are
> ways
> > to
> > > >> achieve that. However they would add extra complexity to Flink,
> which
> > it
> > > >> would be better to avoid.
> > > >>>>>
> > > >>>>> On the other hand, supporting dual pattern: standard partitioning
> > or
> > > >> broadcasting is easy to do, as LatencyMarkers are doing exactly
> that.
> > It
> > > >> would be just a matter of exposing this to the user in some way. So
> > > before
> > > >> we go any further, can you describe your use cases/motivation? Isn’t
> > > mix of
> > > >> standard partitioning and broadcasting enough? Do we need
> > multicasting?
> > > >>>>>
> > > >>>>> Zhu:
> > > >>>>>
> > > >>>>> Could you rephrase your example? I didn’t quite understand it.
> > > >>>>>
> > > >>>>> Piotrek
> > > >>>>>
> > > >>>>> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> > > >> https://issues.apache.org/jira/browse/FLINK-10662> <
> > > >> https://issues.apache.org/jira/browse/FLINK-10662 <
> > > >> https://issues.apache.org/jira/browse/FLINK-10662>>
> > > >>>>>
> > > >>>>> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
> > > >> [hidden email]> <mailto:[hidden email] <mailto:
> > [hidden email]
> > > >>>
> > > >> wrote:
> > > >>>>>
> > > >>>>> Thanks Yun for starting this discussion.
> > > >>>>> I think the multicasting can be very helpful in certain cases.
> > > >>>>>
> > > >>>>> I have received requirements from users that they want to do
> > > broadcast
> > > >>>>> join, while the data set to broadcast is too large to fit in one
> > > task.
> > > >>>>> Thus the requirement turned out to be to support cartesian
> product
> > of
> > > >> 2
> > > >>>>> data set(one of which can be infinite stream).
> > > >>>>> For example, A(parallelism=2) broadcast join B(parallelism=2) in
> > > >> JobVertex
> > > >>>>> C.
> > > >>>>> The idea to is have 4 C subtasks to deal with different
> > combinations
> > > >> of A/B
> > > >>>>> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> > > >>>>> This requires one record to be sent to multiple downstream
> > subtasks,
> > > >> but
> > > >>>>> not to all subtasks.
> > > >>>>>
> > > >>>>> With current interface this is not supported, as one record can
> > only
> > > >> be
> > > >>>>> sent to one subtask, or to all subtasks of a JobVertex.
> > > >>>>> And the user had to split the broadcast data set manually to
> > several
> > > >>>>> different JobVertices, which is hard to maintain and extend.
> > > >>>>>
> > > >>>>> Thanks,
> > > >>>>> Zhu Zhu
> > > >>>>>
> > > >>>>> Yun Gao <[hidden email] <mailto:
> > > >> [hidden email] <mailto:[hidden email]
> >>>
> > > >> 于2019年8月22日周四 下午8:42写道：
> > > >>>>>
> > > >>>>> Hi everyone,
> > > >>>>>    In some scenarios we met a requirement that some operators
> want
> > to
> > > >>>>> send records to theirs downstream operators with an multicast
> > > >> communication
> > > >>>>> pattern. In detail, for some records, the operators want to send
> > them
> > > >>>>> according to the partitioner (for example, Rebalance), and for
> some
> > > >> other
> > > >>>>> records, the operators want to send them to all the connected
> > > >> operators and
> > > >>>>> tasks. Such a communication pattern could be viewed as a kind of
> > > >> multicast:
> > > >>>>> it does not broadcast every record, but some record will indeed
> be
> > > >> sent to
> > > >>>>> multiple downstream operators.
> > > >>>>>
> > > >>>>> However, we found that this kind of communication pattern seems
> > could
> > > >> not
> > > >>>>> be implemented rightly if the operators have multiple consumers
> > with
> > > >>>>> different parallelism, using the customized partitioner. To solve
> > the
> > > >> above
> > > >>>>> problem, we propose to enhance the support for such kind of
> > irregular
> > > >>>>> communication pattern. We think there may be two options:
> > > >>>>>
> > > >>>>>   1. Support a kind of customized operator events, which share
> much
> > > >>>>> similarity with Watermark, and these events can be broadcasted to
> > the
> > > >>>>> downstream operators separately.
> > > >>>>>   2. Let the channel selector supports multicast, and also add
> the
> > > >>>>> separate RecordWriter implementation to avoid impacting the
> > > >> performance of
> > > >>>>> the channel selector that does not need multicast.
> > > >>>>>
> > > >>>>> The problem and options are detailed in
> > > >>>>>
> > > >>
> > >
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > > >> <
> > > >>
> > >
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > > >
> > > >> <
> > > >>
> > >
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > > >> <
> > > >>
> > >
> >
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> > > >>>>
> > > >>>>>
> > > >>>>> We are also wondering if there are other methods to implement
> this
> > > >>>>> requirement with or without changing Runtime. Very thanks for any
> > > >> feedbacks
> > > >>>>> !
> > > >>>>>
> > > >>>>>
> > > >>>>> Best,
> > > >>>>> Yun
> > > >>>>>
> > > >>>>>
> > > >>>>>
> > > >>>>>
> > > >>>>
> > > >>>>
> > > >>>
> > > >>
> > > >>
> > > >>
> > >
> > >
> >
>

[Piotr Nowojski-3]

Hi,

Xiaogang, those things worry me the most.

1. Regarding the broadcasting, doesn’t the BroadcastState [1] cover our issues? Can not we construct a job graph, where one operator has two outputs, one keyed another broadcasted, which are wired together back to the KeyedBroadcastProcessFunction or BroadcastProcessFunction?

2. Multicast on keyed streams, might be done by iterating over all of the keys. However I have a feeling that might not be the feature which distributed cross/theta joins would want, since they would probably need a guarantee to have only a single key per operator instance.

Kurt, by broadcast optimisation do you mean [2]?

I’m not sure if we should split the discussion yet. Most of the changes required by either multicast or broadcast will be in the API/state layers. Runtime changes for broadcast would be almost none (just exposing existing features) and for multicast they shouldn't be huge as well. However maybe we should consider those two things together at the API level, so that we do not make wrong decisions when just looking at the simpler/more narrow broadcast support?

Piotrek

[1] https://ci.apache.org/projects/flink/flink-docs-stable/dev/stream/state/broadcast_state.html <https://ci.apache.org/projects/flink/flink-docs-stable/dev/stream/state/broadcast_state.html>
[2] https://github.com/apache/flink/pull/7713 <https://github.com/apache/flink/pull/7713>

> On 26 Aug 2019, at 09:35, Kurt Young <[hidden email]> wrote:
>
> From SQL's perspective, distributed cross join is a valid feature but not
> very
> urgent. Actually this discuss reminds me about another useful feature
> (sorry
> for the distraction):
>
> when doing broadcast in batch shuffle mode, we can make each producer only
> write one copy of the output data, but not for every consumer. Broadcast
> join
> is much more useful, and this is a very important optimization. Not sure if
> we
> have already consider this.
>
> Best,
> Kurt
>
>
> On Mon, Aug 26, 2019 at 12:16 PM Guowei Ma <[hidden email]> wrote:
>
>> Thanks Yun for bringing up this discussion and very thanks for all the deep
>> thoughts!
>>
>> For now, I think this discussion contains two scenarios: one if for
>> iteration library support and the other is for SQL join support. I think
>> both of the two scenarios are useful but they seem to have different best
>> suitable solutions. For making the discussion more clear, I would suggest
>> to split the discussion into two threads.
>>
>> And I agree with Piotr that it is very tricky that a keyed stream received
>> a "broadcast element". So we may add some new interfaces, which could
>> broadcast or process some special "broadcast event". In that way "broadcast
>> event" will not be sent with the normal process.
>>
>> Best,
>> Guowei
>>
>>
>> SHI Xiaogang <[hidden email]> 于2019年8月26日周一 上午9:27写道：
>>
>>> Hi all,
>>>
>>> I also think that multicasting is a necessity in Flink, but more details
>>> are needed to be considered.
>>>
>>> Currently network is tightly coupled with states in Flink to achieve
>>> automatic scaling. We can only access keyed states in keyed streams and
>>> operator states in all streams.
>>> In the concrete example of theta-joins implemented with mutlticasting,
>> the
>>> following questions exist:
>>>
>>>   - In which type of states will the data be stored? Do we need another
>>>   type of states which is coupled with multicasting streams?
>>>   - How to ensure the consistency between network and states when jobs
>>>   scale out or scale in?
>>>
>>> Regards,
>>> Xiaogang
>>>
>>> Xingcan Cui <[hidden email]> 于2019年8月25日周日 上午10:03写道：
>>>
>>>> Hi all,
>>>>
>>>> Sorry for joining this thread late. Basically, I think enabling
>> multicast
>>>> pattern could be the right direction, but more detailed implementation
>>>> policies need to be discussed.
>>>>
>>>> Two years ago, I filed an issue [1] about the multicast API. However,
>> due
>>>> to some reasons, it was laid aside. After that, when I tried to
>>> cherry-pick
>>>> the change for experimental use, I found the return type of
>>>> `selectChannels()` method had changed from `int[]` to `int`, which
>> makes
>>>> the old implementation not work anymore.
>>>>
>>>> From my side, the multicast has always been used for theta-join. As far
>>> as
>>>> I know, it’s an essential requirement for some sophisticated joining
>>>> algorithms. Until now, the Flink non-equi joins can still only be
>>> executed
>>>> single-threaded. If we'd like to make some improvements on this, we
>>> should
>>>> first take some measures to support multicast pattern.
>>>>
>>>> Best,
>>>> Xingcan
>>>>
>>>> [1] https://issues.apache.org/jira/browse/FLINK-6936
>>>>
>>>>> On Aug 24, 2019, at 5:54 AM, Zhu Zhu <[hidden email]> wrote:
>>>>>
>>>>> Hi Piotr,
>>>>>
>>>>> Thanks for the explanation.
>>>>> Agreed that the broadcastEmit(record) is a better choice for
>>> broadcasting
>>>>> for the iterations.
>>>>> As broadcasting for the iterations is the first motivation, let's
>>> support
>>>>> it first.
>>>>>
>>>>> Thanks,
>>>>> Zhu Zhu
>>>>>
>>>>> Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：
>>>>>
>>>>>>    Hi Piotr,
>>>>>>
>>>>>>     Very thanks for the suggestions!
>>>>>>
>>>>>>    Totally agree with that we could first focus on the broadcast
>>>>>> scenarios and exposing the broadcastEmit method first considering
>> the
>>>>>> semantics and performance.
>>>>>>
>>>>>>    For the keyed stream, I also agree with that broadcasting keyed
>>>>>> records to all the tasks may be confused considering the semantics
>> of
>>>> keyed
>>>>>> partitioner. However, in the iteration case supporting broadcast
>> over
>>>> keyed
>>>>>> partitioner should be required since users may create any subgraph
>> for
>>>> the
>>>>>> iteration body, including the operators with key. I think a possible
>>>>>> solution to this issue is to introduce another data type for
>>>>>> 'broadcastEmit'. For example, for an operator Operator<T>, it may
>>>> broadcast
>>>>>> emit another type E instead of T, and the transmitting E will bypass
>>> the
>>>>>> partitioner and setting keyed context. This should result in the
>>> design
>>>> to
>>>>>> introduce customized operator event (option 1 in the document). The
>>>> cost of
>>>>>> this method is that we need to introduce a new type of StreamElement
>>> and
>>>>>> new interface for this type, but it should be suitable for both
>> keyed
>>> or
>>>>>> non-keyed partitioner.
>>>>>>
>>>>>> Best,
>>>>>> Yun
>>>>>>
>>>>>>
>>>>>>
>>>>>> ------------------------------------------------------------------
>>>>>> From:Piotr Nowojski <[hidden email]>
>>>>>> Send Time:2019 Aug. 23 (Fri.) 22:29
>>>>>> To:Zhu Zhu <[hidden email]>
>>>>>> Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
>>>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
>>>> Pattern
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> If the primary motivation is broadcasting (for the iterations) and
>> we
>>>> have
>>>>>> no immediate need for multicast (cross join), I would prefer to
>> first
>>>>>> expose broadcast via the DataStream API and only later, once we
>>> finally
>>>>>> need it, support multicast. As I wrote, multicast would be more
>>>> challenging
>>>>>> to implement, with more complicated runtime and API. And re-using
>>>> multicast
>>>>>> just to support broadcast doesn’t have much sense:
>>>>>>
>>>>>> 1. It’s a bit obfuscated. It’s easier to understand
>>>>>> collectBroadcast(record) or broadcastEmit(record) compared to some
>>>>>> multicast channel selector that just happens to return all of the
>>>> channels.
>>>>>> 2. There are performance benefits of explicitly calling
>>>>>> `RecordWriter#broadcastEmit`.
>>>>>>
>>>>>>
>>>>>> On a different note, what would be the semantic of such broadcast
>> emit
>>>> on
>>>>>> KeyedStream? Would it be supported? Or would we limit support only
>> to
>>>> the
>>>>>> non-keyed streams?
>>>>>>
>>>>>> Piotrek
>>>>>>
>>>>>>> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
>>>>>>>
>>>>>>> Thanks Piotr,
>>>>>>>
>>>>>>> Users asked for this feature sometimes ago when they migrating
>> batch
>>>>>> jobs to Flink(Blink).
>>>>>>> It's not very urgent as they have taken some workarounds to solve
>>>>>> it.(like partitioning data set to different job vertices)
>>>>>>> So it's fine to not make it top priority.
>>>>>>>
>>>>>>> Anyway, as a commonly known scenario, I think users can benefit
>> from
>>>>>> cross join sooner or later.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Zhu Zhu
>>>>>>>
>>>>>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
>>>>>> 于2019年8月23日周五 下午6:19写道：
>>>>>>> Hi,
>>>>>>>
>>>>>>> Thanks for the answers :) Ok I understand the full picture now. +1
>>> from
>>>>>> my side on solving this issue somehow. But before we start
>> discussing
>>>> how
>>>>>> to solve it one last control question:
>>>>>>>
>>>>>>> I guess this multicast is intended to be used in blink planner,
>>> right?
>>>>>> Assuming that we implement the multicast support now, when would it
>> be
>>>> used
>>>>>> by the blink? I would like to avoid a scenario, where we implement
>> an
>>>>>> unused feature and we keep maintaining it for a long period of time.
>>>>>>>
>>>>>>> Piotrek
>>>>>>>
>>>>>>> PS, try to include motivating examples, including concrete ones in
>>> the
>>>>>> proposals/design docs, for example in the very first paragraph.
>>>> Especially
>>>>>> if it’s a commonly known feature like cross join :)
>>>>>>>
>>>>>>>> On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
>>>>>> wrote:
>>>>>>>>
>>>>>>>>   Hi Piotr,
>>>>>>>>
>>>>>>>>      Thanks a lot for sharing the thoughts!
>>>>>>>>
>>>>>>>>      For the iteration, agree with that multicasting is not
>>>>>> necessary. Exploring the broadcast interface to Output of the
>>> operators
>>>> in
>>>>>> some way should also solve this issue, and I think it should be even
>>>> more
>>>>>> convenient to have the broadcast method for the iteration.
>>>>>>>>
>>>>>>>>      Also thanks Zhu Zhu for the cross join case!
>>>>>>>> Best,
>>>>>>>> Yun
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> ------------------------------------------------------------------
>>>>>>>> From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
>>>>>>>> Send Time:2019 Aug. 23 (Fri.) 17:25
>>>>>>>> To:dev <[hidden email] <mailto:[hidden email]>>
>>>>>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
>>>>>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
>>>>>> Pattern
>>>>>>>>
>>>>>>>> Hi Piotr,
>>>>>>>>
>>>>>>>> Yes you are right it's a distributed cross join requirement.
>>>>>>>> Broadcast join can help with cross join cases. But users cannot
>> use
>>> it
>>>>>> if the data set to join is too large to fit into one subtask.
>>>>>>>>
>>>>>>>> Sorry for left some details behind.
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>> Zhu Zhu
>>>>>>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
>>>>>> 于2019年8月23日周五 下午4:57写道：
>>>>>>>> Hi Yun and Zhu Zhu,
>>>>>>>>
>>>>>>>> Thanks for the more detailed example Zhu Zhu.
>>>>>>>>
>>>>>>>> As far as I understand for the iterations example we do not need
>>>>>> multicasting. Regarding the Join example, I don’t fully understand
>> it.
>>>> The
>>>>>> example that Zhu Zhu presented has a drawback of sending both tables
>>> to
>>>>>> multiple nodes. What’s the benefit of using broadcast join over a
>> hash
>>>> join
>>>>>> in such case? As far as I know, the biggest benefit of using
>> broadcast
>>>> join
>>>>>> instead of hash join is that we can avoid sending the larger table
>>> over
>>>> the
>>>>>> network, because we can perform the join locally. In this example we
>>> are
>>>>>> sending both of the tables to multiple nodes, which should defeat
>> the
>>>>>> purpose.
>>>>>>>>
>>>>>>>> Is it about implementing cross join or near cross joins in a
>>>>>> distributed fashion?
>>>>>>>>
>>>>>>>>> if we introduce a new MulticastRecordWriter
>>>>>>>>
>>>>>>>> That’s one of the solutions. It might have a drawback of 3 class
>>>>>> virtualisation problem (We have RecordWriter and
>> BroadcastRecordWriter
>>>>>> already). With up to two implementations, JVM is able to
>> devirtualise
>>>> the
>>>>>> calls.
>>>>>>>>
>>>>>>>> Previously I was also thinking about just providing two different
>>>>>> ChannelSelector interfaces. One with `int[]` and
>>> `SingleChannelSelector`
>>>>>> with plain `int` and based on that, RecordWriter could perform some
>>>> magic
>>>>>> (worst case scenario `instaceof` checks).
>>>>>>>>
>>>>>>>> Another solution might be to change `ChannelSelector` interface
>> into
>>>>>> an iterator.
>>>>>>>>
>>>>>>>> But let's discuss the details after we agree on implementing this.
>>>>>>>>
>>>>>>>> Piotrek
>>>>>>>>
>>>>>>>>> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
>>>>>> [hidden email]>> wrote:
>>>>>>>>>
>>>>>>>>> Hi Piotr,
>>>>>>>>>
>>>>>>>>>      Thanks a lot for the suggestions!
>>>>>>>>>
>>>>>>>>>      The core motivation of this discussion is to implement a
>> new
>>>>>> iteration library on the DataStream, and it requires to insert
>> special
>>>>>> records in the stream to notify the progress of the iteration. The
>>>>>> mechanism of such records is very similar to the current Watermark,
>>> and
>>>> we
>>>>>> meet the problem of sending normal records according to the
>> partition
>>>>>> (Rebalance, etc..) and also be able to broadcast the inserted
>> progress
>>>>>> records to all the connected records. I have read the notes in the
>>>> google
>>>>>> doc and I totally agree with that exploring the broadcast interface
>> in
>>>>>> RecordWriter in some way is able to solve this issue.
>>>>>>>>>
>>>>>>>>>     Regarding to `int[] ChannelSelector#selectChannels()`, I'm
>>>>>> wondering if we introduce a new MulticastRecordWriter and left the
>>>> current
>>>>>> RecordWriter untouched, could we avoid the performance degradation ?
>>>> Since
>>>>>> with such a modification the normal RecordWriter does not need to
>>>> iterate
>>>>>> the return array by ChannelSelector, and the only difference will be
>>>>>> returning an array instead of an integer, and accessing the first
>>>> element
>>>>>> of the returned array instead of reading the integer directly.
>>>>>>>>>
>>>>>>>>> Best,
>>>>>>>>> Yun
>>>>>>>>>
>>>>>>>>>
>> ------------------------------------------------------------------
>>>>>>>>> From:Piotr Nowojski <[hidden email] <mailto:
>>> [hidden email]
>>>>>>>>
>>>>>>>>> Send Time:2019 Aug. 23 (Fri.) 15:20
>>>>>>>>> To:dev <[hidden email] <mailto:[hidden email]>>
>>>>>>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
>>>>>>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
>>>>>> Pattern
>>>>>>>>>
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> Yun:
>>>>>>>>>
>>>>>>>>> Thanks for proposing the idea. I have checked the document and
>> left
>>>>>> couple of questions there, but it might be better to answer them
>> here.
>>>>>>>>>
>>>>>>>>> What is the exact motivation and what problems do you want to
>>> solve?
>>>>>> We have dropped multicast support from the network stack [1] for two
>>>>>> reasons:
>>>>>>>>> 1. Performance
>>>>>>>>> 2. Code simplicity
>>>>>>>>>
>>>>>>>>> The proposal to re introduce `int[]
>>> ChannelSelector#selectChannels()`
>>>>>> would revert those changes. At that time we were thinking about a
>> way
>>>> how
>>>>>> to keep the multicast support on the network level, while keeping
>> the
>>>>>> performance and simplicity for non multicast cases and there are
>> ways
>>> to
>>>>>> achieve that. However they would add extra complexity to Flink,
>> which
>>> it
>>>>>> would be better to avoid.
>>>>>>>>>
>>>>>>>>> On the other hand, supporting dual pattern: standard partitioning
>>> or
>>>>>> broadcasting is easy to do, as LatencyMarkers are doing exactly
>> that.
>>> It
>>>>>> would be just a matter of exposing this to the user in some way. So
>>>> before
>>>>>> we go any further, can you describe your use cases/motivation? Isn’t
>>>> mix of
>>>>>> standard partitioning and broadcasting enough? Do we need
>>> multicasting?
>>>>>>>>>
>>>>>>>>> Zhu:
>>>>>>>>>
>>>>>>>>> Could you rephrase your example? I didn’t quite understand it.
>>>>>>>>>
>>>>>>>>> Piotrek
>>>>>>>>>
>>>>>>>>> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
>>>>>> https://issues.apache.org/jira/browse/FLINK-10662> <
>>>>>> https://issues.apache.org/jira/browse/FLINK-10662 <
>>>>>> https://issues.apache.org/jira/browse/FLINK-10662>>
>>>>>>>>>
>>>>>>>>> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
>>>>>> [hidden email]> <mailto:[hidden email] <mailto:
>>> [hidden email]
>>>>>>>
>>>>>> wrote:
>>>>>>>>>
>>>>>>>>> Thanks Yun for starting this discussion.
>>>>>>>>> I think the multicasting can be very helpful in certain cases.
>>>>>>>>>
>>>>>>>>> I have received requirements from users that they want to do
>>>> broadcast
>>>>>>>>> join, while the data set to broadcast is too large to fit in one
>>>> task.
>>>>>>>>> Thus the requirement turned out to be to support cartesian
>> product
>>> of
>>>>>> 2
>>>>>>>>> data set(one of which can be infinite stream).
>>>>>>>>> For example, A(parallelism=2) broadcast join B(parallelism=2) in
>>>>>> JobVertex
>>>>>>>>> C.
>>>>>>>>> The idea to is have 4 C subtasks to deal with different
>>> combinations
>>>>>> of A/B
>>>>>>>>> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
>>>>>>>>> This requires one record to be sent to multiple downstream
>>> subtasks,
>>>>>> but
>>>>>>>>> not to all subtasks.
>>>>>>>>>
>>>>>>>>> With current interface this is not supported, as one record can
>>> only
>>>>>> be
>>>>>>>>> sent to one subtask, or to all subtasks of a JobVertex.
>>>>>>>>> And the user had to split the broadcast data set manually to
>>> several
>>>>>>>>> different JobVertices, which is hard to maintain and extend.
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Zhu Zhu
>>>>>>>>>
>>>>>>>>> Yun Gao <[hidden email] <mailto:
>>>>>> [hidden email] <mailto:[hidden email]
>>>>>
>>>>>> 于2019年8月22日周四 下午8:42写道：
>>>>>>>>>
>>>>>>>>> Hi everyone,
>>>>>>>>>   In some scenarios we met a requirement that some operators
>> want
>>> to
>>>>>>>>> send records to theirs downstream operators with an multicast
>>>>>> communication
>>>>>>>>> pattern. In detail, for some records, the operators want to send
>>> them
>>>>>>>>> according to the partitioner (for example, Rebalance), and for
>> some
>>>>>> other
>>>>>>>>> records, the operators want to send them to all the connected
>>>>>> operators and
>>>>>>>>> tasks. Such a communication pattern could be viewed as a kind of
>>>>>> multicast:
>>>>>>>>> it does not broadcast every record, but some record will indeed
>> be
>>>>>> sent to
>>>>>>>>> multiple downstream operators.
>>>>>>>>>
>>>>>>>>> However, we found that this kind of communication pattern seems
>>> could
>>>>>> not
>>>>>>>>> be implemented rightly if the operators have multiple consumers
>>> with
>>>>>>>>> different parallelism, using the customized partitioner. To solve
>>> the
>>>>>> above
>>>>>>>>> problem, we propose to enhance the support for such kind of
>>> irregular
>>>>>>>>> communication pattern. We think there may be two options:
>>>>>>>>>
>>>>>>>>>  1. Support a kind of customized operator events, which share
>> much
>>>>>>>>> similarity with Watermark, and these events can be broadcasted to
>>> the
>>>>>>>>> downstream operators separately.
>>>>>>>>>  2. Let the channel selector supports multicast, and also add
>> the
>>>>>>>>> separate RecordWriter implementation to avoid impacting the
>>>>>> performance of
>>>>>>>>> the channel selector that does not need multicast.
>>>>>>>>>
>>>>>>>>> The problem and options are detailed in
>>>>>>>>>
>>>>>>
>>>>
>>>
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>>>>>> <
>>>>>>
>>>>
>>>
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>>>>>
>>>>>> <
>>>>>>
>>>>
>>>
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>>>>>> <
>>>>>>
>>>>
>>>
>> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
>>>>>>>>
>>>>>>>>>
>>>>>>>>> We are also wondering if there are other methods to implement
>> this
>>>>>>>>> requirement with or without changing Runtime. Very thanks for any
>>>>>> feedbacks
>>>>>>>>> !
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Best,
>>>>>>>>> Yun
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>
>>>>
>>>
>>

[Kurt Young]

Yes, glad to see that there is already a PR for such optimization.

Best,
Kurt


On Mon, Aug 26, 2019 at 6:59 PM Piotr Nowojski <[hidden email]> wrote:

> Hi,
>
> Xiaogang, those things worry me the most.
>
> 1. Regarding the broadcasting, doesn’t the BroadcastState [1] cover our
> issues? Can not we construct a job graph, where one operator has two
> outputs, one keyed another broadcasted, which are wired together back to
> the KeyedBroadcastProcessFunction or BroadcastProcessFunction?
>
> 2. Multicast on keyed streams, might be done by iterating over all of the
> keys. However I have a feeling that might not be the feature which
> distributed cross/theta joins would want, since they would probably need a
> guarantee to have only a single key per operator instance.
>
> Kurt, by broadcast optimisation do you mean [2]?
>
> I’m not sure if we should split the discussion yet. Most of the changes
> required by either multicast or broadcast will be in the API/state layers.
> Runtime changes for broadcast would be almost none (just exposing existing
> features) and for multicast they shouldn't be huge as well. However maybe
> we should consider those two things together at the API level, so that we
> do not make wrong decisions when just looking at the simpler/more narrow
> broadcast support?
>
> Piotrek
>
> [1]
> https://ci.apache.org/projects/flink/flink-docs-stable/dev/stream/state/broadcast_state.html
> <
> https://ci.apache.org/projects/flink/flink-docs-stable/dev/stream/state/broadcast_state.html
> >
> [2] https://github.com/apache/flink/pull/7713 <
> https://github.com/apache/flink/pull/7713>
>
> > On 26 Aug 2019, at 09:35, Kurt Young <[hidden email]> wrote:
> >
> > From SQL's perspective, distributed cross join is a valid feature but not
> > very
> > urgent. Actually this discuss reminds me about another useful feature
> > (sorry
> > for the distraction):
> >
> > when doing broadcast in batch shuffle mode, we can make each producer
> only
> > write one copy of the output data, but not for every consumer. Broadcast
> > join
> > is much more useful, and this is a very important optimization. Not sure
> if
> > we
> > have already consider this.
> >
> > Best,
> > Kurt
> >
> >
> > On Mon, Aug 26, 2019 at 12:16 PM Guowei Ma <[hidden email]> wrote:
> >
> >> Thanks Yun for bringing up this discussion and very thanks for all the
> deep
> >> thoughts!
> >>
> >> For now, I think this discussion contains two scenarios: one if for
> >> iteration library support and the other is for SQL join support. I think
> >> both of the two scenarios are useful but they seem to have different
> best
> >> suitable solutions. For making the discussion more clear, I would
> suggest
> >> to split the discussion into two threads.
> >>
> >> And I agree with Piotr that it is very tricky that a keyed stream
> received
> >> a "broadcast element". So we may add some new interfaces, which could
> >> broadcast or process some special "broadcast event". In that way
> "broadcast
> >> event" will not be sent with the normal process.
> >>
> >> Best,
> >> Guowei
> >>
> >>
> >> SHI Xiaogang <[hidden email]> 于2019年8月26日周一 上午9:27写道：
> >>
> >>> Hi all,
> >>>
> >>> I also think that multicasting is a necessity in Flink, but more
> details
> >>> are needed to be considered.
> >>>
> >>> Currently network is tightly coupled with states in Flink to achieve
> >>> automatic scaling. We can only access keyed states in keyed streams and
> >>> operator states in all streams.
> >>> In the concrete example of theta-joins implemented with mutlticasting,
> >> the
> >>> following questions exist:
> >>>
> >>>   - In which type of states will the data be stored? Do we need another
> >>>   type of states which is coupled with multicasting streams?
> >>>   - How to ensure the consistency between network and states when jobs
> >>>   scale out or scale in?
> >>>
> >>> Regards,
> >>> Xiaogang
> >>>
> >>> Xingcan Cui <[hidden email]> 于2019年8月25日周日 上午10:03写道：
> >>>
> >>>> Hi all,
> >>>>
> >>>> Sorry for joining this thread late. Basically, I think enabling
> >> multicast
> >>>> pattern could be the right direction, but more detailed implementation
> >>>> policies need to be discussed.
> >>>>
> >>>> Two years ago, I filed an issue [1] about the multicast API. However,
> >> due
> >>>> to some reasons, it was laid aside. After that, when I tried to
> >>> cherry-pick
> >>>> the change for experimental use, I found the return type of
> >>>> `selectChannels()` method had changed from `int[]` to `int`, which
> >> makes
> >>>> the old implementation not work anymore.
> >>>>
> >>>> From my side, the multicast has always been used for theta-join. As
> far
> >>> as
> >>>> I know, it’s an essential requirement for some sophisticated joining
> >>>> algorithms. Until now, the Flink non-equi joins can still only be
> >>> executed
> >>>> single-threaded. If we'd like to make some improvements on this, we
> >>> should
> >>>> first take some measures to support multicast pattern.
> >>>>
> >>>> Best,
> >>>> Xingcan
> >>>>
> >>>> [1] https://issues.apache.org/jira/browse/FLINK-6936
> >>>>
> >>>>> On Aug 24, 2019, at 5:54 AM, Zhu Zhu <[hidden email]> wrote:
> >>>>>
> >>>>> Hi Piotr,
> >>>>>
> >>>>> Thanks for the explanation.
> >>>>> Agreed that the broadcastEmit(record) is a better choice for
> >>> broadcasting
> >>>>> for the iterations.
> >>>>> As broadcasting for the iterations is the first motivation, let's
> >>> support
> >>>>> it first.
> >>>>>
> >>>>> Thanks,
> >>>>> Zhu Zhu
> >>>>>
> >>>>> Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：
> >>>>>
> >>>>>>    Hi Piotr,
> >>>>>>
> >>>>>>     Very thanks for the suggestions!
> >>>>>>
> >>>>>>    Totally agree with that we could first focus on the broadcast
> >>>>>> scenarios and exposing the broadcastEmit method first considering
> >> the
> >>>>>> semantics and performance.
> >>>>>>
> >>>>>>    For the keyed stream, I also agree with that broadcasting keyed
> >>>>>> records to all the tasks may be confused considering the semantics
> >> of
> >>>> keyed
> >>>>>> partitioner. However, in the iteration case supporting broadcast
> >> over
> >>>> keyed
> >>>>>> partitioner should be required since users may create any subgraph
> >> for
> >>>> the
> >>>>>> iteration body, including the operators with key. I think a possible
> >>>>>> solution to this issue is to introduce another data type for
> >>>>>> 'broadcastEmit'. For example, for an operator Operator<T>, it may
> >>>> broadcast
> >>>>>> emit another type E instead of T, and the transmitting E will bypass
> >>> the
> >>>>>> partitioner and setting keyed context. This should result in the
> >>> design
> >>>> to
> >>>>>> introduce customized operator event (option 1 in the document). The
> >>>> cost of
> >>>>>> this method is that we need to introduce a new type of StreamElement
> >>> and
> >>>>>> new interface for this type, but it should be suitable for both
> >> keyed
> >>> or
> >>>>>> non-keyed partitioner.
> >>>>>>
> >>>>>> Best,
> >>>>>> Yun
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> ------------------------------------------------------------------
> >>>>>> From:Piotr Nowojski <[hidden email]>
> >>>>>> Send Time:2019 Aug. 23 (Fri.) 22:29
> >>>>>> To:Zhu Zhu <[hidden email]>
> >>>>>> Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
> >>>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> >>>> Pattern
> >>>>>>
> >>>>>> Hi,
> >>>>>>
> >>>>>> If the primary motivation is broadcasting (for the iterations) and
> >> we
> >>>> have
> >>>>>> no immediate need for multicast (cross join), I would prefer to
> >> first
> >>>>>> expose broadcast via the DataStream API and only later, once we
> >>> finally
> >>>>>> need it, support multicast. As I wrote, multicast would be more
> >>>> challenging
> >>>>>> to implement, with more complicated runtime and API. And re-using
> >>>> multicast
> >>>>>> just to support broadcast doesn’t have much sense:
> >>>>>>
> >>>>>> 1. It’s a bit obfuscated. It’s easier to understand
> >>>>>> collectBroadcast(record) or broadcastEmit(record) compared to some
> >>>>>> multicast channel selector that just happens to return all of the
> >>>> channels.
> >>>>>> 2. There are performance benefits of explicitly calling
> >>>>>> `RecordWriter#broadcastEmit`.
> >>>>>>
> >>>>>>
> >>>>>> On a different note, what would be the semantic of such broadcast
> >> emit
> >>>> on
> >>>>>> KeyedStream? Would it be supported? Or would we limit support only
> >> to
> >>>> the
> >>>>>> non-keyed streams?
> >>>>>>
> >>>>>> Piotrek
> >>>>>>
> >>>>>>> On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:
> >>>>>>>
> >>>>>>> Thanks Piotr,
> >>>>>>>
> >>>>>>> Users asked for this feature sometimes ago when they migrating
> >> batch
> >>>>>> jobs to Flink(Blink).
> >>>>>>> It's not very urgent as they have taken some workarounds to solve
> >>>>>> it.(like partitioning data set to different job vertices)
> >>>>>>> So it's fine to not make it top priority.
> >>>>>>>
> >>>>>>> Anyway, as a commonly known scenario, I think users can benefit
> >> from
> >>>>>> cross join sooner or later.
> >>>>>>>
> >>>>>>> Thanks,
> >>>>>>> Zhu Zhu
> >>>>>>>
> >>>>>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> >>>>>> 于2019年8月23日周五 下午6:19写道：
> >>>>>>> Hi,
> >>>>>>>
> >>>>>>> Thanks for the answers :) Ok I understand the full picture now. +1
> >>> from
> >>>>>> my side on solving this issue somehow. But before we start
> >> discussing
> >>>> how
> >>>>>> to solve it one last control question:
> >>>>>>>
> >>>>>>> I guess this multicast is intended to be used in blink planner,
> >>> right?
> >>>>>> Assuming that we implement the multicast support now, when would it
> >> be
> >>>> used
> >>>>>> by the blink? I would like to avoid a scenario, where we implement
> >> an
> >>>>>> unused feature and we keep maintaining it for a long period of time.
> >>>>>>>
> >>>>>>> Piotrek
> >>>>>>>
> >>>>>>> PS, try to include motivating examples, including concrete ones in
> >>> the
> >>>>>> proposals/design docs, for example in the very first paragraph.
> >>>> Especially
> >>>>>> if it’s a commonly known feature like cross join :)
> >>>>>>>
> >>>>>>>> On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
> >>>>>> wrote:
> >>>>>>>>
> >>>>>>>>   Hi Piotr,
> >>>>>>>>
> >>>>>>>>      Thanks a lot for sharing the thoughts!
> >>>>>>>>
> >>>>>>>>      For the iteration, agree with that multicasting is not
> >>>>>> necessary. Exploring the broadcast interface to Output of the
> >>> operators
> >>>> in
> >>>>>> some way should also solve this issue, and I think it should be even
> >>>> more
> >>>>>> convenient to have the broadcast method for the iteration.
> >>>>>>>>
> >>>>>>>>      Also thanks Zhu Zhu for the cross join case!
> >>>>>>>> Best,
> >>>>>>>> Yun
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> ------------------------------------------------------------------
> >>>>>>>> From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
> >>>>>>>> Send Time:2019 Aug. 23 (Fri.) 17:25
> >>>>>>>> To:dev <[hidden email] <mailto:[hidden email]>>
> >>>>>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> >>>>>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> >>>>>> Pattern
> >>>>>>>>
> >>>>>>>> Hi Piotr,
> >>>>>>>>
> >>>>>>>> Yes you are right it's a distributed cross join requirement.
> >>>>>>>> Broadcast join can help with cross join cases. But users cannot
> >> use
> >>> it
> >>>>>> if the data set to join is too large to fit into one subtask.
> >>>>>>>>
> >>>>>>>> Sorry for left some details behind.
> >>>>>>>>
> >>>>>>>> Thanks,
> >>>>>>>> Zhu Zhu
> >>>>>>>> Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
> >>>>>> 于2019年8月23日周五 下午4:57写道：
> >>>>>>>> Hi Yun and Zhu Zhu,
> >>>>>>>>
> >>>>>>>> Thanks for the more detailed example Zhu Zhu.
> >>>>>>>>
> >>>>>>>> As far as I understand for the iterations example we do not need
> >>>>>> multicasting. Regarding the Join example, I don’t fully understand
> >> it.
> >>>> The
> >>>>>> example that Zhu Zhu presented has a drawback of sending both tables
> >>> to
> >>>>>> multiple nodes. What’s the benefit of using broadcast join over a
> >> hash
> >>>> join
> >>>>>> in such case? As far as I know, the biggest benefit of using
> >> broadcast
> >>>> join
> >>>>>> instead of hash join is that we can avoid sending the larger table
> >>> over
> >>>> the
> >>>>>> network, because we can perform the join locally. In this example we
> >>> are
> >>>>>> sending both of the tables to multiple nodes, which should defeat
> >> the
> >>>>>> purpose.
> >>>>>>>>
> >>>>>>>> Is it about implementing cross join or near cross joins in a
> >>>>>> distributed fashion?
> >>>>>>>>
> >>>>>>>>> if we introduce a new MulticastRecordWriter
> >>>>>>>>
> >>>>>>>> That’s one of the solutions. It might have a drawback of 3 class
> >>>>>> virtualisation problem (We have RecordWriter and
> >> BroadcastRecordWriter
> >>>>>> already). With up to two implementations, JVM is able to
> >> devirtualise
> >>>> the
> >>>>>> calls.
> >>>>>>>>
> >>>>>>>> Previously I was also thinking about just providing two different
> >>>>>> ChannelSelector interfaces. One with `int[]` and
> >>> `SingleChannelSelector`
> >>>>>> with plain `int` and based on that, RecordWriter could perform some
> >>>> magic
> >>>>>> (worst case scenario `instaceof` checks).
> >>>>>>>>
> >>>>>>>> Another solution might be to change `ChannelSelector` interface
> >> into
> >>>>>> an iterator.
> >>>>>>>>
> >>>>>>>> But let's discuss the details after we agree on implementing this.
> >>>>>>>>
> >>>>>>>> Piotrek
> >>>>>>>>
> >>>>>>>>> On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
> >>>>>> [hidden email]>> wrote:
> >>>>>>>>>
> >>>>>>>>> Hi Piotr,
> >>>>>>>>>
> >>>>>>>>>      Thanks a lot for the suggestions!
> >>>>>>>>>
> >>>>>>>>>      The core motivation of this discussion is to implement a
> >> new
> >>>>>> iteration library on the DataStream, and it requires to insert
> >> special
> >>>>>> records in the stream to notify the progress of the iteration. The
> >>>>>> mechanism of such records is very similar to the current Watermark,
> >>> and
> >>>> we
> >>>>>> meet the problem of sending normal records according to the
> >> partition
> >>>>>> (Rebalance, etc..) and also be able to broadcast the inserted
> >> progress
> >>>>>> records to all the connected records. I have read the notes in the
> >>>> google
> >>>>>> doc and I totally agree with that exploring the broadcast interface
> >> in
> >>>>>> RecordWriter in some way is able to solve this issue.
> >>>>>>>>>
> >>>>>>>>>     Regarding to `int[] ChannelSelector#selectChannels()`, I'm
> >>>>>> wondering if we introduce a new MulticastRecordWriter and left the
> >>>> current
> >>>>>> RecordWriter untouched, could we avoid the performance degradation ?
> >>>> Since
> >>>>>> with such a modification the normal RecordWriter does not need to
> >>>> iterate
> >>>>>> the return array by ChannelSelector, and the only difference will be
> >>>>>> returning an array instead of an integer, and accessing the first
> >>>> element
> >>>>>> of the returned array instead of reading the integer directly.
> >>>>>>>>>
> >>>>>>>>> Best,
> >>>>>>>>> Yun
> >>>>>>>>>
> >>>>>>>>>
> >> ------------------------------------------------------------------
> >>>>>>>>> From:Piotr Nowojski <[hidden email] <mailto:
> >>> [hidden email]
> >>>>>>>>
> >>>>>>>>> Send Time:2019 Aug. 23 (Fri.) 15:20
> >>>>>>>>> To:dev <[hidden email] <mailto:[hidden email]>>
> >>>>>>>>> Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
> >>>>>>>>> Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
> >>>>>> Pattern
> >>>>>>>>>
> >>>>>>>>> Hi,
> >>>>>>>>>
> >>>>>>>>> Yun:
> >>>>>>>>>
> >>>>>>>>> Thanks for proposing the idea. I have checked the document and
> >> left
> >>>>>> couple of questions there, but it might be better to answer them
> >> here.
> >>>>>>>>>
> >>>>>>>>> What is the exact motivation and what problems do you want to
> >>> solve?
> >>>>>> We have dropped multicast support from the network stack [1] for two
> >>>>>> reasons:
> >>>>>>>>> 1. Performance
> >>>>>>>>> 2. Code simplicity
> >>>>>>>>>
> >>>>>>>>> The proposal to re introduce `int[]
> >>> ChannelSelector#selectChannels()`
> >>>>>> would revert those changes. At that time we were thinking about a
> >> way
> >>>> how
> >>>>>> to keep the multicast support on the network level, while keeping
> >> the
> >>>>>> performance and simplicity for non multicast cases and there are
> >> ways
> >>> to
> >>>>>> achieve that. However they would add extra complexity to Flink,
> >> which
> >>> it
> >>>>>> would be better to avoid.
> >>>>>>>>>
> >>>>>>>>> On the other hand, supporting dual pattern: standard partitioning
> >>> or
> >>>>>> broadcasting is easy to do, as LatencyMarkers are doing exactly
> >> that.
> >>> It
> >>>>>> would be just a matter of exposing this to the user in some way. So
> >>>> before
> >>>>>> we go any further, can you describe your use cases/motivation? Isn’t
> >>>> mix of
> >>>>>> standard partitioning and broadcasting enough? Do we need
> >>> multicasting?
> >>>>>>>>>
> >>>>>>>>> Zhu:
> >>>>>>>>>
> >>>>>>>>> Could you rephrase your example? I didn’t quite understand it.
> >>>>>>>>>
> >>>>>>>>> Piotrek
> >>>>>>>>>
> >>>>>>>>> [1] https://issues.apache.org/jira/browse/FLINK-10662 <
> >>>>>> https://issues.apache.org/jira/browse/FLINK-10662> <
> >>>>>> https://issues.apache.org/jira/browse/FLINK-10662 <
> >>>>>> https://issues.apache.org/jira/browse/FLINK-10662>>
> >>>>>>>>>
> >>>>>>>>> On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
> >>>>>> [hidden email]> <mailto:[hidden email] <mailto:
> >>> [hidden email]
> >>>>>>>
> >>>>>> wrote:
> >>>>>>>>>
> >>>>>>>>> Thanks Yun for starting this discussion.
> >>>>>>>>> I think the multicasting can be very helpful in certain cases.
> >>>>>>>>>
> >>>>>>>>> I have received requirements from users that they want to do
> >>>> broadcast
> >>>>>>>>> join, while the data set to broadcast is too large to fit in one
> >>>> task.
> >>>>>>>>> Thus the requirement turned out to be to support cartesian
> >> product
> >>> of
> >>>>>> 2
> >>>>>>>>> data set(one of which can be infinite stream).
> >>>>>>>>> For example, A(parallelism=2) broadcast join B(parallelism=2) in
> >>>>>> JobVertex
> >>>>>>>>> C.
> >>>>>>>>> The idea to is have 4 C subtasks to deal with different
> >>> combinations
> >>>>>> of A/B
> >>>>>>>>> partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
> >>>>>>>>> This requires one record to be sent to multiple downstream
> >>> subtasks,
> >>>>>> but
> >>>>>>>>> not to all subtasks.
> >>>>>>>>>
> >>>>>>>>> With current interface this is not supported, as one record can
> >>> only
> >>>>>> be
> >>>>>>>>> sent to one subtask, or to all subtasks of a JobVertex.
> >>>>>>>>> And the user had to split the broadcast data set manually to
> >>> several
> >>>>>>>>> different JobVertices, which is hard to maintain and extend.
> >>>>>>>>>
> >>>>>>>>> Thanks,
> >>>>>>>>> Zhu Zhu
> >>>>>>>>>
> >>>>>>>>> Yun Gao <[hidden email] <mailto:
> >>>>>> [hidden email] <mailto:[hidden email]
> >>>>>
> >>>>>> 于2019年8月22日周四 下午8:42写道：
> >>>>>>>>>
> >>>>>>>>> Hi everyone,
> >>>>>>>>>   In some scenarios we met a requirement that some operators
> >> want
> >>> to
> >>>>>>>>> send records to theirs downstream operators with an multicast
> >>>>>> communication
> >>>>>>>>> pattern. In detail, for some records, the operators want to send
> >>> them
> >>>>>>>>> according to the partitioner (for example, Rebalance), and for
> >> some
> >>>>>> other
> >>>>>>>>> records, the operators want to send them to all the connected
> >>>>>> operators and
> >>>>>>>>> tasks. Such a communication pattern could be viewed as a kind of
> >>>>>> multicast:
> >>>>>>>>> it does not broadcast every record, but some record will indeed
> >> be
> >>>>>> sent to
> >>>>>>>>> multiple downstream operators.
> >>>>>>>>>
> >>>>>>>>> However, we found that this kind of communication pattern seems
> >>> could
> >>>>>> not
> >>>>>>>>> be implemented rightly if the operators have multiple consumers
> >>> with
> >>>>>>>>> different parallelism, using the customized partitioner. To solve
> >>> the
> >>>>>> above
> >>>>>>>>> problem, we propose to enhance the support for such kind of
> >>> irregular
> >>>>>>>>> communication pattern. We think there may be two options:
> >>>>>>>>>
> >>>>>>>>>  1. Support a kind of customized operator events, which share
> >> much
> >>>>>>>>> similarity with Watermark, and these events can be broadcasted to
> >>> the
> >>>>>>>>> downstream operators separately.
> >>>>>>>>>  2. Let the channel selector supports multicast, and also add
> >> the
> >>>>>>>>> separate RecordWriter implementation to avoid impacting the
> >>>>>> performance of
> >>>>>>>>> the channel selector that does not need multicast.
> >>>>>>>>>
> >>>>>>>>> The problem and options are detailed in
> >>>>>>>>>
> >>>>>>
> >>>>
> >>>
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >>>>>> <
> >>>>>>
> >>>>
> >>>
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >>>>>
> >>>>>> <
> >>>>>>
> >>>>
> >>>
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >>>>>> <
> >>>>>>
> >>>>
> >>>
> >>
> https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
> >>>>>>>>
> >>>>>>>>>
> >>>>>>>>> We are also wondering if there are other methods to implement
> >> this
> >>>>>>>>> requirement with or without changing Runtime. Very thanks for any
> >>>>>> feedbacks
> >>>>>>>>> !
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Best,
> >>>>>>>>> Yun
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>
> >>>>
> >>>
> >>
>
>

[Yun Gao]

Hi,

Very thanks for all the points raised !

@Piotr For using another edge to broadcast the event, I think it may not be able to address the iteration case. The primary problem is that with  two edges we cannot ensure the order of records. However, In the iteration case, the broadcasted event is used to mark the progress of the iteration and it works like watermark, thus its position relative to the normal records can not change.
And @Piotr, @Xiaogang, for the requirements on the state, I think different options seems vary. The first option is to allow Operator<T> to broadcast a separate event and have a separate process method for this event. To be detail, we may add a new type of StreamElement called Event and allow Operator<T> to broadcastEmit Event. Then in the received side, we could add a new `processEvent` method to the (Keyed)ProcessFunction. Similar to the broadcast side of KeyedBroadcastProcessFunction, in this new method users cannot access keyed state with specific key, but can register a state function to touch all the elements in the keyed state. This option needs to modify the runtime to support the new type of StreamElement, but it does not affect the semantics of states and thus it has no requirements on state.
The second option is to allow Operator<T> to broadcastEmit T and in the receiver side, user can process the broadcast element with the existing process method. This option is consistent with the OperatorState, but for keyedState we may send a record to tasks that do not containing the corresponding keyed state, thus it should require some changes on the State.
The third option is to support the generic Multicast. For keyedState it also meets the problem of inconsistency between network partitioner and keyed state partitioner, and if we want to rely on it to implement the non-key join, it should be also meet the problem of cannot control the partitioning of operator state. Therefore, it should also require some changes on the State.
Then for the different scenarios proposed, the iteration case in fact requires exactly the ability to broadcast a different event type. In the iteration the fields of the progress event are in fact different from that of normal records. It does not contain actual value but contains some fields for the downstream operators to align the events and track the progress. Therefore, broadcasting a different event type is able to solve the iteration case without the requirements on the state. Besides, allowing the operator to broadcast a separate event may also facilitate some other user cases, for example, users may notify the downstream operators to change logic if some patterns are matched. The notification might be different from the normal records and users do not need to uniform them with a wrapper type manually if the operators are able to broadcast a separate event. However, it truly cannot address the non-key join scenarios.
Since allowing broadcasting a separate event seems to be able to serve as a standalone functionality, and it does not require change on the state, I am thinking that is it possible for us to partition to multiple steps and supports broadcasting events first ? At the same time we could also continue working on other options to support more scenarios like non-key join and they seems to requires more thoughts.

Best,
Yun



------------------------------------------------------------------
From:Piotr Nowojski <[hidden email]>
Send Time:2019 Aug. 26 (Mon.) 18:59
To:dev <[hidden email]>
Cc:Yun Gao <[hidden email]>
Subject:Re: [DISCUSS] Enhance Support for Multicast Communication Pattern

Hi,

Xiaogang, those things worry me the most.
1. Regarding the broadcasting, doesn’t the BroadcastState [1] cover our issues? Can not we construct a job graph, where one operator has two outputs, one keyed another broadcasted, which are wired together back to the KeyedBroadcastProcessFunction or BroadcastProcessFunction?

2. Multicast on keyed streams, might be done by iterating over all of the keys. However I have a feeling that might not be the feature which distributed cross/theta joins would want, since they would probably need a guarantee to have only a single key per operator instance.

Kurt, by broadcast optimisation do you mean [2]?

I’m not sure if we should split the discussion yet. Most of the changes required by either multicast or broadcast will be in the API/state layers. Runtime changes for broadcast would be almost none (just exposing existing features) and for multicast they shouldn't be huge as well. However maybe we should consider those two things together at the API level, so that we do not make wrong decisions when just looking at the simpler/more narrow broadcast support?

Piotrek

[1] https://ci.apache.org/projects/flink/flink-docs-stable/dev/stream/state/broadcast_state.html
[2] https://github.com/apache/flink/pull/7713


On 26 Aug 2019, at 09:35, Kurt Young <[hidden email]> wrote:
From SQL's perspective, distributed cross join is a valid feature but not
very
urgent. Actually this discuss reminds me about another useful feature
(sorry
for the distraction):

when doing broadcast in batch shuffle mode, we can make each producer only
write one copy of the output data, but not for every consumer. Broadcast
join
is much more useful, and this is a very important optimization. Not sure if
we
have already consider this.

Best,
Kurt


On Mon, Aug 26, 2019 at 12:16 PM Guowei Ma <[hidden email]> wrote:

Thanks Yun for bringing up this discussion and very thanks for all the deep
thoughts!

For now, I think this discussion contains two scenarios: one if for
iteration library support and the other is for SQL join support. I think
both of the two scenarios are useful but they seem to have different best
suitable solutions. For making the discussion more clear, I would suggest
to split the discussion into two threads.

And I agree with Piotr that it is very tricky that a keyed stream received
a "broadcast element". So we may add some new interfaces, which could
broadcast or process some special "broadcast event". In that way "broadcast
event" will not be sent with the normal process.

Best,
Guowei


SHI Xiaogang <[hidden email]> 于2019年8月26日周一 上午9:27写道：

Hi all,

I also think that multicasting is a necessity in Flink, but more details
are needed to be considered.

Currently network is tightly coupled with states in Flink to achieve
automatic scaling. We can only access keyed states in keyed streams and
operator states in all streams.
In the concrete example of theta-joins implemented with mutlticasting,
the
following questions exist:

   - In which type of states will the data be stored? Do we need another
   type of states which is coupled with multicasting streams?
   - How to ensure the consistency between network and states when jobs
   scale out or scale in?

Regards,
Xiaogang

Xingcan Cui <[hidden email]> 于2019年8月25日周日 上午10:03写道：

Hi all,

Sorry for joining this thread late. Basically, I think enabling
multicast
pattern could be the right direction, but more detailed implementation
policies need to be discussed.

Two years ago, I filed an issue [1] about the multicast API. However,
due
to some reasons, it was laid aside. After that, when I tried to
cherry-pick
the change for experimental use, I found the return type of
`selectChannels()` method had changed from `int[]` to `int`, which
makes
the old implementation not work anymore.

From my side, the multicast has always been used for theta-join. As far
as
I know, it’s an essential requirement for some sophisticated joining
algorithms. Until now, the Flink non-equi joins can still only be
executed
single-threaded. If we'd like to make some improvements on this, we
should
first take some measures to support multicast pattern.

Best,
Xingcan

[1] https://issues.apache.org/jira/browse/FLINK-6936

On Aug 24, 2019, at 5:54 AM, Zhu Zhu <[hidden email]> wrote:

Hi Piotr,

Thanks for the explanation.
Agreed that the broadcastEmit(record) is a better choice for
broadcasting
for the iterations.
As broadcasting for the iterations is the first motivation, let's
support
it first.

Thanks,
Zhu Zhu

Yun Gao <[hidden email]> 于2019年8月23日周五 下午11:56写道：

    Hi Piotr,

     Very thanks for the suggestions!

    Totally agree with that we could first focus on the broadcast
scenarios and exposing the broadcastEmit method first considering
the
semantics and performance.

    For the keyed stream, I also agree with that broadcasting keyed
records to all the tasks may be confused considering the semantics
of
keyed
partitioner. However, in the iteration case supporting broadcast
over
keyed
partitioner should be required since users may create any subgraph
for
the
iteration body, including the operators with key. I think a possible
solution to this issue is to introduce another data type for
'broadcastEmit'. For example, for an operator Operator<T>, it may
broadcast
emit another type E instead of T, and the transmitting E will bypass
the
partitioner and setting keyed context. This should result in the
design
to
introduce customized operator event (option 1 in the document). The
cost of
this method is that we need to introduce a new type of StreamElement
and
new interface for this type, but it should be suitable for both
keyed
or
non-keyed partitioner.

Best,
Yun



------------------------------------------------------------------
From:Piotr Nowojski <[hidden email]>
Send Time:2019 Aug. 23 (Fri.) 22:29
To:Zhu Zhu <[hidden email]>
Cc:dev <[hidden email]>; Yun Gao <[hidden email]>
Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
Pattern

Hi,

If the primary motivation is broadcasting (for the iterations) and
we
have
no immediate need for multicast (cross join), I would prefer to
first
expose broadcast via the DataStream API and only later, once we
finally
need it, support multicast. As I wrote, multicast would be more
challenging
to implement, with more complicated runtime and API. And re-using
multicast
just to support broadcast doesn’t have much sense:

1. It’s a bit obfuscated. It’s easier to understand
collectBroadcast(record) or broadcastEmit(record) compared to some
multicast channel selector that just happens to return all of the
channels.
2. There are performance benefits of explicitly calling
`RecordWriter#broadcastEmit`.


On a different note, what would be the semantic of such broadcast
emit
on
KeyedStream? Would it be supported? Or would we limit support only
to
the
non-keyed streams?

Piotrek

On 23 Aug 2019, at 12:48, Zhu Zhu <[hidden email]> wrote:

Thanks Piotr,

Users asked for this feature sometimes ago when they migrating
batch
jobs to Flink(Blink).
It's not very urgent as they have taken some workarounds to solve
it.(like partitioning data set to different job vertices)
So it's fine to not make it top priority.

Anyway, as a commonly known scenario, I think users can benefit
from
cross join sooner or later.

Thanks,
Zhu Zhu

Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
于2019年8月23日周五 下午6:19写道：
Hi,

Thanks for the answers :) Ok I understand the full picture now. +1
from
my side on solving this issue somehow. But before we start
discussing
how
to solve it one last control question:

I guess this multicast is intended to be used in blink planner,
right?
Assuming that we implement the multicast support now, when would it
be
used
by the blink? I would like to avoid a scenario, where we implement
an
unused feature and we keep maintaining it for a long period of time.

Piotrek

PS, try to include motivating examples, including concrete ones in
the
proposals/design docs, for example in the very first paragraph.
Especially
if it’s a commonly known feature like cross join :)

On 23 Aug 2019, at 11:38, Yun Gao <[hidden email]>
wrote:

   Hi Piotr,

      Thanks a lot for sharing the thoughts!

      For the iteration, agree with that multicasting is not
necessary. Exploring the broadcast interface to Output of the
operators
in
some way should also solve this issue, and I think it should be even
more
convenient to have the broadcast method for the iteration.

      Also thanks Zhu Zhu for the cross join case!
Best,
 Yun



------------------------------------------------------------------
From:Zhu Zhu <[hidden email] <mailto:[hidden email]>>
Send Time:2019 Aug. 23 (Fri.) 17:25
To:dev <[hidden email] <mailto:[hidden email]>>
Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
Pattern

Hi Piotr,

Yes you are right it's a distributed cross join requirement.
Broadcast join can help with cross join cases. But users cannot
use
it
if the data set to join is too large to fit into one subtask.

Sorry for left some details behind.

Thanks,
Zhu Zhu
Piotr Nowojski <[hidden email] <mailto:[hidden email]>>
于2019年8月23日周五 下午4:57写道：
Hi Yun and Zhu Zhu,

Thanks for the more detailed example Zhu Zhu.

As far as I understand for the iterations example we do not need
multicasting. Regarding the Join example, I don’t fully understand
it.
The
example that Zhu Zhu presented has a drawback of sending both tables
to
multiple nodes. What’s the benefit of using broadcast join over a
hash
join
in such case? As far as I know, the biggest benefit of using
broadcast
join
instead of hash join is that we can avoid sending the larger table
over
the
network, because we can perform the join locally. In this example we
are
sending both of the tables to multiple nodes, which should defeat
the
purpose.

Is it about implementing cross join or near cross joins in a
distributed fashion?

if we introduce a new MulticastRecordWriter

That’s one of the solutions. It might have a drawback of 3 class
virtualisation problem (We have RecordWriter and
BroadcastRecordWriter
already). With up to two implementations, JVM is able to
devirtualise
the
calls.

Previously I was also thinking about just providing two different
ChannelSelector interfaces. One with `int[]` and
`SingleChannelSelector`
with plain `int` and based on that, RecordWriter could perform some
magic
(worst case scenario `instaceof` checks).

Another solution might be to change `ChannelSelector` interface
into
an iterator.

But let's discuss the details after we agree on implementing this.

Piotrek

On 23 Aug 2019, at 10:20, Yun Gao <[hidden email] <mailto:
[hidden email]>> wrote:

 Hi Piotr,

      Thanks a lot for the suggestions!

      The core motivation of this discussion is to implement a
new
iteration library on the DataStream, and it requires to insert
special
records in the stream to notify the progress of the iteration. The
mechanism of such records is very similar to the current Watermark,
and
we
meet the problem of sending normal records according to the
partition
(Rebalance, etc..) and also be able to broadcast the inserted
progress
records to all the connected records. I have read the notes in the
google
doc and I totally agree with that exploring the broadcast interface
in
RecordWriter in some way is able to solve this issue.

     Regarding to `int[] ChannelSelector#selectChannels()`, I'm
wondering if we introduce a new MulticastRecordWriter and left the
current
RecordWriter untouched, could we avoid the performance degradation ?
Since
with such a modification the normal RecordWriter does not need to
iterate
the return array by ChannelSelector, and the only difference will be
returning an array instead of an integer, and accessing the first
element
of the returned array instead of reading the integer directly.

Best,
Yun


------------------------------------------------------------------
From:Piotr Nowojski <[hidden email] <mailto:
[hidden email]

Send Time:2019 Aug. 23 (Fri.) 15:20
To:dev <[hidden email] <mailto:[hidden email]>>
Cc:Yun Gao <[hidden email] <mailto:[hidden email]>>
Subject:Re: [DISCUSS] Enhance Support for Multicast Communication
Pattern

Hi,

Yun:

Thanks for proposing the idea. I have checked the document and
left
couple of questions there, but it might be better to answer them
here.

What is the exact motivation and what problems do you want to
solve?
We have dropped multicast support from the network stack [1] for two
reasons:
1. Performance
2. Code simplicity

The proposal to re introduce `int[]
ChannelSelector#selectChannels()`
would revert those changes. At that time we were thinking about a
way
how
to keep the multicast support on the network level, while keeping
the
performance and simplicity for non multicast cases and there are
ways
to
achieve that. However they would add extra complexity to Flink,
which
it
would be better to avoid.

On the other hand, supporting dual pattern: standard partitioning
or
broadcasting is easy to do, as LatencyMarkers are doing exactly
that.
It
would be just a matter of exposing this to the user in some way. So
before
we go any further, can you describe your use cases/motivation? Isn’t
mix of
standard partitioning and broadcasting enough? Do we need
multicasting?

Zhu:

Could you rephrase your example? I didn’t quite understand it.

Piotrek

[1] https://issues.apache.org/jira/browse/FLINK-10662 <
https://issues.apache.org/jira/browse/FLINK-10662> <
https://issues.apache.org/jira/browse/FLINK-10662 <
https://issues.apache.org/jira/browse/FLINK-10662>>

On 23 Aug 2019, at 09:17, Zhu Zhu <[hidden email] <mailto:
[hidden email]> <mailto:[hidden email] <mailto:
[hidden email]

wrote:

Thanks Yun for starting this discussion.
I think the multicasting can be very helpful in certain cases.

I have received requirements from users that they want to do
broadcast
join, while the data set to broadcast is too large to fit in one
task.
Thus the requirement turned out to be to support cartesian
product
of
2
data set(one of which can be infinite stream).
For example, A(parallelism=2) broadcast join B(parallelism=2) in
JobVertex
C.
The idea to is have 4 C subtasks to deal with different
combinations
of A/B
partitions, like C1(A1,B1), C2(A1,B2), C3(A2,B1), C4(A2,B2).
This requires one record to be sent to multiple downstream
subtasks,
but
not to all subtasks.

With current interface this is not supported, as one record can
only
be
sent to one subtask, or to all subtasks of a JobVertex.
And the user had to split the broadcast data set manually to
several
different JobVertices, which is hard to maintain and extend.

Thanks,
Zhu Zhu

Yun Gao <[hidden email] <mailto:
[hidden email] <mailto:[hidden email]

于2019年8月22日周四 下午8:42写道：

Hi everyone,
   In some scenarios we met a requirement that some operators
want
to
send records to theirs downstream operators with an multicast
communication
pattern. In detail, for some records, the operators want to send
them
according to the partitioner (for example, Rebalance), and for
some
other
records, the operators want to send them to all the connected
operators and
tasks. Such a communication pattern could be viewed as a kind of
multicast:
it does not broadcast every record, but some record will indeed
be
sent to
multiple downstream operators.

However, we found that this kind of communication pattern seems
could
not
be implemented rightly if the operators have multiple consumers
with
different parallelism, using the customized partitioner. To solve
the
above
problem, we propose to enhance the support for such kind of
irregular
communication pattern. We think there may be two options:

  1. Support a kind of customized operator events, which share
much
similarity with Watermark, and these events can be broadcasted to
the
downstream operators separately.
  2. Let the channel selector supports multicast, and also add
the
separate RecordWriter implementation to avoid impacting the
performance of
the channel selector that does not need multicast.

The problem and options are detailed in




https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
<



https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing

<



https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing
<



https://docs.google.com/document/d/1npi5c_SeP68KuT2lNdKd8G7toGR_lxQCGOnZm_hVMks/edit?usp=sharing


We are also wondering if there are other methods to implement
this
requirement with or without changing Runtime. Very thanks for any
feedbacks
!


Best,
Yun