[
http://jira.amdatu.org/jira/browse/AMDATU-301?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=10904#comment-10904
]
Ivo Ladage - van Doorn commented on AMDATU-301:
-----------------------------------------------
The major cause of the performance issue was the fact that tenant resolving was
really expensive (each request ended up in a Cassandra query). A tenant cache
will resolve this issue, which is covered by
http://jira.amdatu.org/jira/browse/AMDATU-304.
This also explains why this issue was not found with the dispatcher performance
integration test created before, as this test doesn't use the cassandra tenant
store but an in-memory tenant store, which acts as a cache itself.
> Major increase in response times (factor 46) and decrease of throughput (98%)
> in 0.1.0 compared to 0.0.6
> --------------------------------------------------------------------------------------------------------
>
> Key: AMDATU-301
> URL: http://jira.amdatu.org/jira/browse/AMDATU-301
> Project: Amdatu
> Issue Type: Bug
> Components: Amdatu Web
> Affects Versions: 0.1.0
> Reporter: Ivo Ladage - van Doorn
> Assignee: Bram de Kruijff
> Priority: Blocker
> Fix For: 0.1.0
>
>
> Running the jmeter dashboard scripts from /etc/jmeter in Amdatu 0.0.6, Amdatu
> 0.1.0 prior to the dispatcher merge and Amdatu 0.1.0 latest version shows a
> significant performance issue in the latest 0.1.0 version. Response times go
> up by a factor 46 and throughput is decreased by 98%, compared to 0.0.6.
> To verify this, simply download and run JMeter and run the jmeter scripts
> 'Amdatu dashboard - 0.0.6.jmx' on 0.0.6 and 'Amdatu dashboard - 0.1.0.jmx' on
> 0.1.0.
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
