Mailing-List: contact commits-help@cassandra.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@cassandra.apache.org
Date: Fri, 28 Mar 2014 18:23:20 +0000 (UTC)
From: "Benedict (JIRA)" <jira@apache.org>
To: commits@cassandra.apache.org
Message-ID: <JIRA.12702474.1395261829216.26274.1396031000142@arcas>
In-Reply-To: <JIRA.12702474.1395261829216@arcas>
References: <JIRA.12702474.1395261829216@arcas>
Subject: [jira] [Commented] (CASSANDRA-6890) Remove mmap'ed i/o from read
 path
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/CASSANDRA-6890?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13951179#comment-13951179 ] 

Benedict commented on CASSANDRA-6890:
-------------------------------------

FTR, with compression enabled it will only use mmapped IO for reading the index file, not the compressed data.

I'm not tremendously surprised that we see a drop in throughput vs mmapped files, as we compress 64K chunks, so we have to read a minimum of 64K (+ readahead), and also decompress it for every row we lookup, which in this case is only ~200 bytes. I would expect the performance difference to be smaller for larger rows. Tuning performance for standard IO is trickier for the user, and we don't currently do much to help them.

Personally, I am -1 on removing mmapped IO as an option until we have an in-process uncompressed page cache that mitigates these problems.


> Remove mmap'ed i/o from read path
> ---------------------------------
>
>                 Key: CASSANDRA-6890
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6890
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Joshua McKenzie
>            Assignee: Joshua McKenzie
>            Priority: Minor
>             Fix For: 3.0
>
>         Attachments: mmap_gc.jpg, mmap_jstat.txt, mmap_perf.txt, nommap_gc.jpg, nommap_jstat.txt
>
>
> Since we actively unmap unreferenced SSTR's and also copy data out of those readers on the read path, the current memory mapped i/o is a lot of complexity for very little payoff.  Clean out the mmapp'ed i/o on the read path.


--
This message was sent by Atlassian JIRA
(v6.2#6252)