[
https://issues.apache.org/jira/browse/CASSANDRA-3578?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13835923#comment-13835923
]
Jonathan Ellis edited comment on CASSANDRA-3578 at 12/1/13 2:12 AM:
--------------------------------------------------------------------
bq. We point next to next, which is absolutely correct
I understand what sync is doing; my point is that xxxxSyncMarkerPosition always
points to the *last* offset before which the contents are synced, which we'll
revisit when we sync again / write the next marker (referred to appropriately
as {{nextMarker}} in the code). Your own comments refer to "last" or
"previous" in multiple places which speaks to how difficult it is to avoid
thinking of it that way. :)
(N.B. This is also clear in the usage of isFullySynced -- intuitively, we want
to know if "the *last* place we synced was the end of the buffer" which is
unclear if the variable is named *next*.)
was (Author: jbellis):
bq. We point next to next, which is absolutely correct
I understand what sync is doing; my point is that xxxxSyncMarkerPosition always
points to the *last* offset before which the contents are synced, which we'll
revisit when we sync again / write the next marker (referred to appropriately
as {{nextMarker}} in the code). Your own comments refer to "last" or
"previous" in multiple places which speaks to how difficult it is to avoid
thinking of it that way. :)
(N.B. This is also clear in the usage of isFullySynced -- intuitively, we want
to know if "the *last8 place we synced was the end of the buffer" which is
unclear if the variable is named *next*.)
> Multithreaded commitlog
> -----------------------
>
> Key: CASSANDRA-3578
> URL: https://issues.apache.org/jira/browse/CASSANDRA-3578
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Jonathan Ellis
> Assignee: Benedict
> Priority: Minor
> Labels: performance
> Attachments: 0001-CASSANDRA-3578.patch, ComitlogStress.java,
> Current-CL.png, Multi-Threded-CL.png, TestEA.java, latency.svg, oprate.svg,
> parallel_commit_log_2.patch
>
>
> Brian Aker pointed out a while ago that allowing multiple threads to modify
> the commitlog simultaneously (reserving space for each with a CAS first, the
> way we do in the SlabAllocator.Region.allocate) can improve performance,
> since you're not bottlenecking on a single thread to do all the copying and
> CRC computation.
> Now that we use mmap'd CommitLog segments (CASSANDRA-3411) this becomes
> doable.
> (moved from CASSANDRA-622, which was getting a bit muddled.)
--
This message was sent by Atlassian JIRA
(v6.1#6144)
