[jboss-jira] [JBoss JIRA] Commented: (BPEL-282) Engine runs process instances in serie instead of in parallel

[Alejandro Guizar (JIRA)]

    [ http://jira.jboss.com/jira/browse/BPEL-282?page=comments#action_12384976 ] 
            
Alejandro Guizar commented on BPEL-282:
---------------------------------------

Currently, all start message listeners run on a single JMS session. Since JMS delivers messages one at a time per session, this feature may cause the perceived serial execution. I have made slight modifications which allow each start listener to run in a separate JMS session. This should enable parallel execution as far as the JMS connection and, ultimately, the available CPU cycles permit it.

Once I complete the QA procedure I will upload a new version of the jBPM BPEL binaries to the repository and mark the issue as resolved.

> Engine runs process instances in serie instead of in parallel
> -------------------------------------------------------------
>
>                 Key: BPEL-282
>                 URL: http://jira.jboss.com/jira/browse/BPEL-282
>             Project: JBoss jBPM BPEL
>          Issue Type: Bug
>      Security Level: Public(Everyone can see) 
>          Components: Engine
>    Affects Versions: jBPM BPEL 1.1 beta 3
>         Environment: RHEL
>            Reporter: Santiago Erquicia
>         Assigned To: Alejandro Guizar
>
> We are stress testing a BPEL process and we found out that the engine runs all the requests it receives in serie. That means that if there are two or more concurrent users of the system, one has to wait till the first process instance stops at some point (wait state) so the other request can be processed.
> On our test environment, we are calling a web service which is also calling another web service. The are connection problems with the second one, so from time to time the web service just times out. This means that a concurrent user has to wait for that time out, and whatever is left of that process instance, in order to be processed.
> On our JMeter test, the throughput keeps stable even if there number of request is bigger than the throughput of the BPEL engine. The result is that the time for response of each process request increases indefinitely if the usage is bigger than the throughput. This obviously doesn't scale at all for the processing requirement we are considering.

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