accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Luke Brassard (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-1083) add concurrency to HDFS write-ahead log
Date Thu, 07 Mar 2013 22:16:12 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-1083?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13596465#comment-13596465
] 

Luke Brassard commented on ACCUMULO-1083:
-----------------------------------------

I just finished running a handful of tests against different versions of Accumulo with walogs
on and off. We are running against hadoop-1.0.3.16 and zookeeper-3.3.4.15.

Here are the results:

||version||table.walog.enabled||MB/s||
|accumulo-1.4.2|true|6.691|
|accumulo-1.4.2|false|23.583|
|accumulo-1.5.0-SNAPSHOT|true|5.734|
|accumulo-1.5.0-SNAPSHOT |false|27.321|

The penalties for walogs are about 72% for 1.4.2 and 79% for 1.5.0. So, it's not an extreme
decrease in performance, but there is certainly a dip. We suspect that the performance hit
has to do with the Hadoop bottleneck described above. 

I have not modified the replication factor, but may run those tests too if time permits.
                
> add concurrency to HDFS write-ahead log
> ---------------------------------------
>
>                 Key: ACCUMULO-1083
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-1083
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: tserver
>            Reporter: Adam Fuchs
>             Fix For: 1.6.0
>
>         Attachments: walog-performance.jpg
>
>
> When running tablet servers on beefy nodes (lots of disks), the write-ahead log can be
a serious bottleneck. Today we ran a test of 1.5-SNAPSHOT on an 8-node (plus a master node)
cluster in which the nodes had 32 cores and 15 drives each. Running with write-ahead log off
resulted in a >4x performance improvement sustained over a long period.
> I believe the culprit is that the WAL is only using one file at a time per tablet server,
which means HDFS is only appending to one drive (plus replicas). If we increase the number
of concurrent WAL files supported on a tablet server we could probably drastically improve
the performance on systems with many disks. As it stands, I believe Accumulo is significantly
more optimized for a larger number of smaller nodes (3-4 drives).

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message