Hi, In storm we could do thing like:
TopologyBuilder builder = new TopologyBuilder(); builder.setSpout("spout", new NumberSpout()); builder.setBolt(“mybolt", new Mybolt()) .shuffleGrouping("spout") .shuffleGrouping(“mybolt", “iterativeStream"); It means that after one operation there are two output streams. One of them will be a input of the operation. Seemly, Flink streaming also supports this feature - https://ci.apache.org/projects/flink/flink-docs-release-0.10/apis/streaming_guide.html#iterations. My question is does Spark streaming support this feature also? If yes, how? I couldn’t find it from the Internet. Thanks for your help. Jun