Impact on performance while using session service
I am working on a comparison of performance when directly accessing a service S1 and when going through session.
To directly access the service S1, I created a proxy and ran a method of the service.
To access the service through session service, created the required session and then sent the message to Session service to route to the required service S1.(The session creation time is not considered in total time taken)
The results I obtained indicates that the processing through session service takes approx. 5-7 times the time taken while accessing the service directly.
I have made all the changes as specified in the CSF Documentation for performance improvements.
Are there any others things which can be considered to improve the performance.
Ashish
Hi Ashish,
Thanks for your note - here are some analysis questions to help clarify the tests you are performing:
(1) Are you messuring the time it takes for CreateSession and message routing?
(2) Does the Session manifest contain a persona participant?
(3) How many participants are in the session?
(4) How did you build the S1 service? Does it use the WSE 3.0 library? Does it use the CSFSDK? Is it an ASMX web service?
Many Thanks, Ashish!!
-Rob.
Hi Rob,
Please find the answers to your questions:
(1) Are you messuring the time it takes for CreateSession and message routing?
- The time being measured is only for message routing. The total time measured is (Time of receiving response - Time of sending request to session service)
(2) Does the Session manifest contain a persona participant?
- Yes, the service S1 I am trying uses a custom token manager to validate user from a database.
(3) How many participants are in the session?
- 4(including one for persona)
(4) How did you build the S1 service? Does it use the WSE 3.0 library? Does it use the CSFSDK? Is it an ASMX web service?
- The service S1 is a web service created using the WES Toolkit(Well Enables Service Toolkit) with some customization done for functionality and security changes.
Let me know if more information is required.
Ashish
Hi Ashish,
Thanks for your note - a couple of follow-up questions for you:
(a) Have you checked that WSE tracing is disabled - set to false - in Session's web.config file?
(b) Are you using Kerberos Security in the call to S1, or UsernameToken?
(c) Have you enabled the WSDL caching feature in Session? We recommend doing this for performance testing...
Many Thanks, Ashish!!
-Rob.
Hi Rob,
pleaes find my answers inline.
(a) Have you checked that WSE tracing is disabled - set to false - in Session's web.config file?
Its currently set to true for debugging purposes
(b) Are you using Kerberos Security in the call to S1, or UsernameToken?
Username Token
(c) Have you enabled the WSDL caching feature in Session? We recommend doing this for performance testing...
Yes, its configured with default setting.
Also I rerun the test by setting the tracing to False and there was massive improvement with the performance. Now the timings coming are almost the same as a direct call to the service using proxy.
Regards,
Ashish
