[ https://issues.apache.org/jira/browse/FLINK-2548?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14704730#comment-14704730 ]
Gabor Gevay commented on FLINK-2548: ------------------------------------ {quote} CoGrouping with the solution set is a different runtime operator than the regular CoGroup. {quote} I see, thanks! I think this should be documented in the iterations programming guide, and in the javadoc for coGroup and getSolutionSet. {quote} (2) To have an iterator for the "message" sending, you can group by source vertex. Then you group again by target vertex, to have an iterator for incoming messages (either reduce+join or CoGroup). So you have now join + reduce + reduce + join in teh iteration? {quote} For calling the MessagingFunction, I join the edges with the workset, and then do a reduceGroup by source vertex. This gives me an iterator over the neighbors inside the UDF of the reduceGroup. (Well, almost. That iterator is a little different from what the messaging function expects, so I wrote an iterator adaptor.) See my code \[1\], and here is an execution plan which involves the new code: \[2\]. I think this code is memory-safe in the way you described. \[1\] https://github.com/apache/flink/compare/master...ggevay:gelly-remove-coGroup-first-version \[2\] http://compalg.inf.elte.hu/~ggevay/flink/Plan_join_groupBy.txt > VertexCentricIteration should avoid doing a coGroup with the edges and the > solution set > --------------------------------------------------------------------------------------- > > Key: FLINK-2548 > URL: https://issues.apache.org/jira/browse/FLINK-2548 > Project: Flink > Issue Type: Improvement > Components: Gelly > Affects Versions: 0.9, 0.10 > Reporter: Gabor Gevay > Assignee: Gabor Gevay > > Currently, the performance of vertex centric iteration is suboptimal in those > iterations where the workset is small, because the complexity of one > iteration contains the number of edges and vertices of the graph because of > coGroups: > VertexCentricIteration.buildMessagingFunction does a coGroup between the > edges and the workset, to get the neighbors to the messaging UDF. This is > problematic from a performance point of view, because the coGroup UDF gets > called on all the edge groups, including those that are not getting any > messages. > An analogous problem is present in > VertexCentricIteration.createResultSimpleVertex at the creation of the > updates: a coGroup happens between the messages and the solution set, which > has the number of vertices of the graph included in its complexity. > Both of these coGroups could be avoided by doing a join instead (with the > same keys that the coGroup uses), and then a groupBy. The complexity of these > operations would be dominated by the size of the workset, as opposed to the > number of edges or vertices of the graph. The joins should have the edges and > the solution set at the build side to achieve this complexity. (They will not > be rebuilt at every iteration.) > I made some experiments with this, and the initial results seem promising. On > some workloads, this achieves a 2 times speedup, because later iterations > often have quite small worksets, and these get a huge speedup from this. -- This message was sent by Atlassian JIRA (v6.3.4#6332)