[
https://issues.apache.org/jira/browse/FLINK-8172?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16272496#comment-16272496
]
ASF GitHub Bot commented on FLINK-8172:
---------------------------------------
Github user pnowojski commented on the issue:
https://github.com/apache/flink/pull/5104
2 times improvement is for network stack only. With other things in play
(like even simple operators) it will be smaller.
As we discussed offline, I didn't want tightening up the
connection/dependencies of `RecordSerializer`. `Buffer` is a much simpler
dependency compare to `BufferProvider`. Think how more difficult would it be to
test it with `BufferProvider`. However I was thinking about refactoring it in
another direction, creating kind of `write-only` `@NotThreadSafe`
`BufferBuilder` class.
I have create a separate issue for that:
https://issues.apache.org/jira/browse/FLINK-8178
> Remove unnecessary synchronisation in RecordSerializer
> ------------------------------------------------------
>
> Key: FLINK-8172
> URL: https://issues.apache.org/jira/browse/FLINK-8172
> Project: Flink
> Issue Type: Improvement
> Components: Network
> Affects Versions: 1.4.0, 1.3.2
> Reporter: Piotr Nowojski
> Assignee: Piotr Nowojski
> Fix For: 1.5.0
>
>
> While writing the records, RecordSerializer is the only owner of the `Buffer`
> into which data are written. Yet we are synchronisation twice per record
> while accessing MemorySegment. Removing this synchronisation speeds up the
> Network throughput in point to point benchmark by a factor of two (from
> ~12500records/ms up to 23000 records/ms).
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
