[jbosscache-issues] [JBoss JIRA] Updated: (JBCACHE-1566) Bad performance when bigger objects are used

tbech (JIRA) jira-events at lists.jboss.org
Tue Jan 19 06:02:55 EST 2010


     [ https://jira.jboss.org/jira/browse/JBCACHE-1566?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

tbech updated JBCACHE-1566:
---------------------------

    Attachment: SimpleTest.java


> Bad performance when bigger objects are used
> --------------------------------------------
>
>                 Key: JBCACHE-1566
>                 URL: https://jira.jboss.org/jira/browse/JBCACHE-1566
>             Project: JBoss Cache
>          Issue Type: Bug
>      Security Level: Public(Everyone can see) 
>            Reporter: tbech
>            Assignee: Manik Surtani
>         Attachments: SimpleTest.java
>
>
> I've used benchmark framework to test jbosscache with various sizes of the objects in the case.
> Suprisingly in my tests, there was degradation of performance for bigger objects.
> I've modifed StringTest.java in the way, that is PUT into cache random strings (with defined size).
> For 50 character strings, cache performed 1800 PUTs/s for 4 nodes. But for 500 character strings, it was 750 and 2000 character strings it was just 50. 
> In localMode (one-node), it run the same undepending on the string size.
> I've double checked configuration, network, multicast - it looks ok.
> (I've applied http://www.redhat.com/docs/en-US/JBoss_Enterprise_Application_Platform/5.0.0/html/Administration_And_Configuration_Guide/jgroups-perf-udpbuffer.html, but it didn't change anything).
> I've noticed one more strange behaviour: the random string I've generated with the function:
> 	   return RandomStringUtils.random(2000, true, true);
> (RandomStringUtils comes from jakarta commons).
> It generates strings consisting of standard characters and digits.
> However when the generated string consists of all unicode characters
> 	return RandomStringUtils.random(500, true, true);
> and the size was 500, the performance of the test was similar as 2000 standard characters string. Is it some problem with serialization of objects? (As far as I remember in JBoss you use custom serializer).
> I've repeated those tests with Ehcache 1.7.2 and there was no significant drop of performance for 50/2000 chars strings.
> I attach modified SimpleTest.java.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

        


More information about the jbosscache-issues mailing list