[
https://issues.apache.org/jira/browse/CASSANDRA-20486?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dmitry Konstantinov updated CASSANDRA-20486:
--------------------------------------------
Summary: Apply memtable sharding to local write request thread pool (was:
Apply memtable sharding to write request thread pool)
> Apply memtable sharding to local write request thread pool
> ----------------------------------------------------------
>
> Key: CASSANDRA-20486
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20486
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Local/Memtable
> Reporter: Dmitry Konstantinov
> Priority: Normal
> Attachments: image-2025-03-26-21-03-21-342.png
>
>
> An idea to evaluate: our memtables use sharding approach with a single writer
> guarded by lock or use optimistic concurrency with retries in case of
> contention, so when we have a high rate of writes we spend time/CPU for lock
> parking/unparking or by contending on per partition optimistic locking.
> !image-2025-03-26-21-03-21-342.png|width=300!
> A possible way to reduce or even almost avoid such overheads can be applying
> the same kind of sharding for write request threads.
> It is not a real TPC but an easy step towards this direction.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
