accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eric Newton (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-883) evaluate different thrift components for performance
Date Wed, 28 Nov 2012 16:16:58 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-883?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13505607#comment-13505607
] 

Eric Newton commented on ACCUMULO-883:
--------------------------------------

OK, you're right.  Mutation does a bit of it's own serialization. Here are the numbers for
10K calls to continueMultiScan which returns 200 KeyValues:

||{{TServer}}||{{TProtocol}}||Times|| || || Average||
|{{THaHs}}|{{TCompactProtocol}}|5614|3569|3409|4197|
| |{{TBinaryProtocol}}|5624|3260|3977|4287|
|{{ThreadedSelector}}|{{TCompactProtocol}}|5326|2943|3448|3906|
| |{{TBinaryProtocol}}|4729|2963|3200|3630|
|{{TThreadPool}}|{{TCompactProtocol}}|4179|2648|2683|3170|
| |{{TBinaryProtocol}}|4534|3741|3577|3950|

Hmm... looks like TCompactProtocol does help, even when using the very-fast loopback device,
which is what the test uses.
                
> evaluate different thrift components for performance
> ----------------------------------------------------
>
>                 Key: ACCUMULO-883
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-883
>             Project: Accumulo
>          Issue Type: Task
>          Components: master, tserver
>            Reporter: Eric Newton
>            Assignee: Eric Newton
>             Fix For: 1.5.0
>
>
> Now that we've upgraded to Thrift 0.9, evaluate the fastest combination of TServer and
TProtocol on typical server hardware (many cores, lots of ram, fast networks).

--
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