[
https://issues.apache.org/jira/browse/CASSANDRA-6809?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14289242#comment-14289242
]
Benedict commented on CASSANDRA-6809:
-------------------------------------
bq. Thank you, I did not realise you are interested in parallelism between
segments only.
Well, I considered that a natural extension, i.e. a follow up ticket. One I
still consider reasonably straight forward to add: a mutator thread can
partition the commit range once it's processed ~1Mb, and simply append the
Callable to a shared queue. The sync thread can then drain this when it decides
to initiate a sync.
bq. I can see that this should work well enough with large sync periods,
including the 10s default.
I'm reasonably confident this will work as well or better for all sync periods.
In particular it better guarantees honouring the sync periods, and is less
likely to encourage random write behaviour. Of course, the main benefit is its
simplicity.
> Compressed Commit Log
> ---------------------
>
> Key: CASSANDRA-6809
> URL: https://issues.apache.org/jira/browse/CASSANDRA-6809
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Benedict
> Assignee: Branimir Lambov
> Priority: Minor
> Labels: performance
> Fix For: 3.0
>
> Attachments: ComitLogStress.java, logtest.txt
>
>
> It seems an unnecessary oversight that we don't compress the commit log.
> Doing so should improve throughput, but some care will need to be taken to
> ensure we use as much of a segment as possible. I propose decoupling the
> writing of the records from the segments. Basically write into a (queue of)
> DirectByteBuffer, and have the sync thread compress, say, ~64K chunks every X
> MB written to the CL (where X is ordinarily CLS size), and then pack as many
> of the compressed chunks into a CLS as possible.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
