[
https://issues.apache.org/jira/browse/CASSANDRA-5549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13874168#comment-13874168
]
Benedict commented on CASSANDRA-5549:
-------------------------------------
bq. WaitQueue has changed enough that git no longer recognizes the new one as
sharing history with the old one. Can you summarize?
Sure:
- It includes the aforementioned improvements with NBQ (including now being
non-blocking, except in the case of signaller/waiter trampling on each other
occasionally in pthread world).
- It abstracts the Signal, so that we can have other implementations: wait for
all, wait for any, for example.
- It supports timing the amount of time spent waiting. I have toyed with
splitting this out into another ticket, but I did it whilst I was going, as I
thought having the amount of time spent waiting on Memtable (or CL) flush would
be a helpful metric.
That said, perhaps we could try splitting this whole thing into a few smaller
tickets, make it more manageable for you to review. I kind of suspect the main
body won't diminish a lot though, which is why I didn't in the first place. But
every little probably helps :-)
> Remove Table.switchLock
> -----------------------
>
> Key: CASSANDRA-5549
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5549
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jonathan Ellis
> Assignee: Benedict
> Labels: performance
> Fix For: 2.1
>
> Attachments: 5549-removed-switchlock.png, 5549-sunnyvale.png
>
>
> As discussed in CASSANDRA-5422, Table.switchLock is a bottleneck on the write
> path. ReentrantReadWriteLock is not lightweight, even if there is no
> contention per se between readers and writers of the lock (in Cassandra,
> memtable updates and switches).
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
