[ 
https://issues.apache.org/jira/browse/CASSANDRA-8897?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14544849#comment-14544849
 ] 

Stefania commented on CASSANDRA-8897:
-------------------------------------

I've restored the CLQ in Ref.State and removed the reference to the channel in 
the RAR constructor. For those cases where we do not have a shared channel I 
created a thin wrapper of the RAR that closes the channel in the close method, 
see the latest commit on my branch. I prefer this to hiding the non-ownership 
of the channel behind an empty SharedCloseableImpl that does not hold a Ref.

I've attached an updated 9240_test_results.txt for the tests on my machine and 
it seems the 8897 ops per second are now slightly better than trunk if 
anything. I've also repeated the tests with Java 8, similar results but 
slightly better than Java 7.

Here are the repeat cstar_perf tests:

- 9240 profile with reduced memory : 
http://cstar.datastax.com/tests/id/c7f1729e-fa9f-11e4-94a5-42010af0688f
- standard 5M stress with 50 threads : 
http://cstar.datastax.com/tests/id/e6bbe33a-fa9f-11e4-94a5-42010af0688f
- standard 15M stress with 50 threads : 
http://cstar.datastax.com/tests/id/6def435a-faa6-11e4-9b87-42010af0688f

Performance looks comparable.

Latest Jenkins results are also still good.

> Remove FileCacheService, instead pooling the buffers
> ----------------------------------------------------
>
>                 Key: CASSANDRA-8897
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-8897
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Benedict
>            Assignee: Stefania
>             Fix For: 3.x
>
>         Attachments: 9240_test_results.txt, 
> snapshot-1431582436640-cpu-backtraces.png, 
> snapshot-1431582436640-cpu-calltree-compression-8897.nps, 
> snapshot-1431582436640-cpu-calltree-compression-trunk.nps
>
>
> After CASSANDRA-8893, a RAR will be a very lightweight object and will not 
> need caching, so we can eliminate this cache entirely. Instead we should have 
> a pool of buffers that are page-aligned.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to