Tom Cunningham created JBESB-3952: ------------------------------------- Summary: Performance regression with Bpm5Processor Key: JBESB-3952 URL: https://issues.jboss.org/browse/JBESB-3952 Project: JBoss ESB Issue Type: Bug Security Level: Public (Everyone can see) Components: Process flow Affects Versions: 4.11 CP3 Reporter: Tom Cunningham Assignee: Tom Cunningham Fix For: 4.11 CP3 I did a quick test comparing the original code they used (based on the SOA-5.3.0 jBPM5 integration layer) with the new one (based on the SOA-5.3.1 jBPM5 integration layer). The main difference between the two codebases is the session utilization pattern. 5.3.0 uses a session-per-process-instance pattern, 5.3.1 uses a session-per-process-definition pattern. I did a test with a process that takes 4 sec to execute, using a JMS listener with 20 threads of executions. With the 5.3.0. code, it takes 90 sec to execute 380 process instances. With the 5.3.1 code, it takes 27 min to execute 380 process instances. The reason for this huge performance difference is simple: the 5.3.1 uses only one session, and the this session is synchronized (thus single-threaded). So the benefit of using multiple threads of execution in the ESB is lost. -- 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