[
https://issues.apache.org/jira/browse/FLINK-19026?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17201553#comment-17201553
]
Arvid Heise commented on FLINK-19026:
-------------------------------------
Merged to master as
https://github.com/apache/flink/commit/69921a88b9de723207c56d585b34d1bc5c3b67d3
.
> Improve threading model of Unaligner
> ------------------------------------
>
> Key: FLINK-19026
> URL: https://issues.apache.org/jira/browse/FLINK-19026
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Network
> Affects Versions: 1.12.0
> Reporter: Arvid Heise
> Assignee: Arvid Heise
> Priority: Major
> Labels: pull-request-available
>
> Currently, netty threads and task thread are both accessing Unaligner. Netty
> thread access it indirectly through BufferReceivedListener. Thus, we need
> some costly synchronization and weaken the threading model of the Unaligner.
> One possibility would be to introduce priority buffers to the InputGates,
> which overtake other buffers and also are always processable (e.g. they
> bypass non-blocking outputs).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
