Wow! Incredible :-) Can you share more details about the experiments you ran (cluster setup, jobs, etc)?
On Fri, Aug 8, 2014 at 3:53 PM, Gyula Fóra <gyula.f...@gmail.com> wrote: > Hey All, > > Quick weekely update on the streaming project: > > It was a good week we implemented a lot of new features and made > considerable work on the api too. Most notably: > > - Cluster performance was measured against Storm on both simple streaming > wordcount and iterative algorithm (pagerank) and Flink Streaming was about > 4 times faster. (partly because the output buffers) > > - The API now support simple types instead of Tuple1s (so instead of > DataStream<Tuple1<String>> you can use DataStream<String>) > > - The API was updated to match the new function interfaces as in the Main > project, with both the standard and the Rich functions > > - The Directed emit api has been updated to include SplitDataStreams and a > .select(name) method to direct tuples to named outputs. > > - We have added a .groupBy(..) operator to use with .reduce, .batchReduce > and .windowReduce allowing a streaming group-reduce on the whole stream, > sliding batches, and sliding time windows. > > - We have also started refactoring our tests to run in much less time to > avoid travis errors :) > > You can check all these changes on git of course: > > https://github.com/mbalassi/incubator-flink/commits/streaming-ready > > > Cheers, > Gyula >