[
https://issues.apache.org/jira/browse/CASSANDRA-9533?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14658924#comment-14658924
]
Ariel Weisberg commented on CASSANDRA-9533:
-------------------------------------------
Is this updating a shared cache line for every single write that flows through
C* even though the signaling only needs to occur infrequently? Can you
reformulate this so a write only occurs if the thread being notified is
actually asleep?
Looking at things it seems a little barn door, but I can attest to the big
difference it makes. The CL I wrote was very sensitive to submission cost and
topped out around 250k operations/second and that number did vary a lot
depending on how much shared state had to be mutated as part of submission. It
was the one big not split lock in the system.
> Make batch commitlog mode easier to tune
> ----------------------------------------
>
> Key: CASSANDRA-9533
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9533
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Jonathan Ellis
> Assignee: Benedict
> Fix For: 3.x
>
>
> As discussed in CASSANDRA-9504, 2.1 changed commitlog_sync_batch_window_in_ms
> from a maximum time to wait between fsync to the minimum time, so one must be
> very careful to keep it small enough that most writers aren't kept waiting.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
