cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Ellis (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-8075) Make netty pipeline fully async
Date Thu, 09 Oct 2014 16:18:35 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-8075?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14165317#comment-14165317
] 

Jonathan Ellis commented on CASSANDRA-8075:
-------------------------------------------

What is the difference between 5239 and the "fully async" you describe here?

> Make netty pipeline fully async
> -------------------------------
>
>                 Key: CASSANDRA-8075
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-8075
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: T Jake Luciani
>              Labels: performance
>             Fix For: 3.0
>
>
> Related to CASSANDRA-5239.  Netty does much better if we remove the request thread pool
jump. 
> Here is a benchmark of our netty pipeline returning a static value (so no actual cassandra
processing).  When the thread pool is removed we see a dramatic improvement.
> http://cstar.datastax.com/graph?stats=64ed5186-4e51-11e4-a8d5-bc764e04482c&metric=op_rate&operation=2_user&smoothing=1&show_aggregates=true&xmin=0&xmax=228.25&ymin=0&ymax=150217.1
> On my laptop using the async branch can match redis bench on the same machine ~160k ops/sec.



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

Mime
View raw message