[
https://issues.apache.org/jira/browse/CASSANDRA-8692?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14349008#comment-14349008
]
Benedict commented on CASSANDRA-8692:
-------------------------------------
Just as an FYI and a placeholder, since I don't have time to work on it myself
at the moment, but Ariel and I have discussed another strategy that may yield
yet superior performance (more universally) without needing tuning: essentially
one that self tunes, by incrementally increasing or decreasing the coalesce
window based on the most recent changes in throughput, so that it stabilises
near a window that delivers the most yield. i.e. if throughput increases from
the prior interval, it increases the window by a small amount, and vice versa.
It would also train its thresholds and increments based on prior results, but
with tight bounds on resulting effect.
My hope is this would yield current behaviour as a lower bound, but also find a
stable spot for clusters where performance can be improved without any
necessary intervention.
> Coalesce intra-cluster network messages
> ---------------------------------------
>
> Key: CASSANDRA-8692
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8692
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Ariel Weisberg
> Assignee: Ariel Weisberg
> Fix For: 2.1.4
>
> Attachments: batching-benchmark.png
>
>
> While researching CASSANDRA-8457 we found that it is effective and can be
> done without introducing additional latency at low concurrency/throughput.
> The patch from that was used and found to be useful in a real life scenario
> so I propose we implement this in 2.1 in addition to 3.0.
> The change set is a single file and is small enough to be reviewable.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
