You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Pedro Miguel Rainho Chaves (JIRA)" <ji...@apache.org> on 2017/07/11 18:51:00 UTC
[jira] [Created] (FLINK-7157) Provide support for asynchronous
operations over connected streams
Pedro Miguel Rainho Chaves created FLINK-7157:
-------------------------------------------------
Summary: Provide support for asynchronous operations over connected streams
Key: FLINK-7157
URL: https://issues.apache.org/jira/browse/FLINK-7157
Project: Flink
Issue Type: New Feature
Components: DataStream API
Reporter: Pedro Miguel Rainho Chaves
The asynchronous function provides a great way to process events in parallel, which can greatly increase the throughput. However, a user may want to process events in parallel based on a stored state created by another stream. For instance, one could have a stream of scripts (e.g. groovy scripts) and a stream of events. Processing the events by the scripts in a CoFlatMap greatly reduces the performance if those take a long time to execute, specially in cases with a high number of scripts.
One could argue that increasing the parallelism of the CoFlatMap would solve the problem, however it would come at a very high resource cost.
It would be nice to have a *CoAsyncFunction* which would allow one of the streams in a *connectedStream* to be processed in an asynchronous fashion.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)