Flink Benchmarks have been included in BigDataBench

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Flink Benchmarks have been included in BigDataBench

Xinhui Tian
Hello everyone!

Earlier we have posted a topic about the plan of including Flink support in our big data benchmark, which is called BigDataBench. The benchmarks can be found on http://prof.ict.ac.cn/BigDataBench/. Now we have released a new version including Flink benchmarks. It is now a beta version and the benchmarks include basic batch processing like word count, grep, and iteration processing like page rank, delta page rank and KMeans. For better comparison with Spark, we used the KMeans++ algorithm, and also included a KMeans|| version.

We have also released a graph benchmark including five benchmarks on Flink Gelly, Spark GraphX and GraphLab, and we believe such benchmarks may help users to compare those systems in an easy way.

Other benchmarks such as streaming processing are in process.

We also have got some preliminary results about the performance comparison of Hadoop, Spark and Flink, which has been put on website: http://www.bafst.com/top100.html

We hope we can get more suggestions about the performance evaluation and benchmark designs. If you have any advice or question, please post it to our group on LinkedIn: http://www.linkedin.com/groups/6777483?trk=hp-feed-group-name, or contract me directly by sending an email to tianxinhui@ict.ac.cn, we will be very grateful for every advice.

Thanks very much :)
Reply | Threaded
Open this post in threaded view
|

Re: Flink Benchmarks have been included in BigDataBench

Ufuk Celebi-2

> On 18 Dec 2015, at 03:07, Xinhui Tian <[hidden email]> wrote:
>
> Hello everyone!
>
> Earlier we have posted a topic about the plan of including Flink support in
> our big data benchmark, which is called BigDataBench. The benchmarks can be
> found on http://prof.ict.ac.cn/BigDataBench/. Now we have released a new
> version including Flink benchmarks. It is now a beta version and the
> benchmarks include basic batch processing like word count, grep, and
> iteration processing like page rank, delta page rank and KMeans. For better
> comparison with Spark, we used the KMeans++ algorithm, and also included a
> KMeans|| version.
>
> We have also released a graph benchmark including five benchmarks on Flink
> Gelly, Spark GraphX and GraphLab, and we believe such benchmarks may help
> users to compare those systems in an easy way.
>
> Other benchmarks such as streaming processing are in process.
>
> We also have got some preliminary results about the performance comparison
> of Hadoop, Spark and Flink, which has been put on website:
> http://www.bafst.com/top100.html

Thanks for notifying the Flink community!

The prelim results link doesn’t work though.

– Ufuk