Hi
I am running a load test on Glassfish/OpenESB with simple HelloWorld web service invocation from BPLE.
It's simple App which receive a SOAP request, goes to BPLE SE which invokes the Web service running on Glassfish.
JMeter Client (Incoming Request:SOAP) ==> HTTP-BC ==>BPLE SE ==> Invoke in a while loop with 5 iteration==> HTTP-BC ==>Hello Web Service ==>Hello Web Service Response ==> HTTP-BC ==> BPLE Reply ==>HTTP-BC ==>JMeter client)
Here Web Service just returns a static string so no business logic involved.
Above application gives me around 350 TPS (Transactions per Second)
Now if I comment out the Invoke like from BPLE and still execute the BPLE Loop, I get around 1700 TPS which means BPLE execution is not a bottle neck.
I ran another test with JMeter which send direct requests to the Hello Web Service (without going through ESB), I get roughly 2000 TPS.
So it seems Web service invocation within BPLE SE is very expensive and is a bottleneck for performance. I have tried to optimize based on few recommendations so now I get 350 TPS. Earlier it was getting 280 TPS.
Any suggestions how can I improve the performance of web service invocation from BPLE SE?
[Message sent by forum member 'rinku24' (rinku24)]
http://forums.java.net/jive/thread.jspa?messageID=337298