jBPM perfomance testing

[jcaraneda]

I am evaluating jBPM and I have a something questions about it's performance:
1. Are exists perfomance testing?
2. How much process instances I can load in jBPM and how much request rate support good?
3. It has been implemented in great applications with more than 1000 users?
4. Which are the resources that I need to load more than 1000 process instances running?

Thanks for your answers...
JC

[kukeltje]

Hi,

1: Not public ones afaik
2: Process instances that do nothing are just saved in the database, so the only limit there is the db size. The jBPM core can support lots of requests per second. Tuning the system to fit your environment is in most cases the best idea. Comparing different implementations with different requirements can lead to wrong conclusion.
3: Depends on what you mean by more than 1000 users. 1000 simultaneous requests? No, no but not many systems can do that. If you mean 1000 users on the system doing normal operations, so wait-time in between etc... that should be no problem for the core engine. You will need a good hardware and tuned appserver and db however.
4: Again, more than 1000 instances doing nothing is just space in the DB. It realy depends on the real use of the application.

[jcaraneda]

1: Not public ones afaik
Do you have any idea or estimation about the perfomance?

[kukeltje]

Yes, my idea is that the performance is good ;-)

The company I work for is starting to use jBPM, within several weeks I think I can publish some numbers. Maybe Tom or Koen can post some numbers, but I know customers are not keen on this and environments can differ so much that sometimes figures are not representative.

[jcaraneda]

I doing a stress test over websale example. Soon I can publish numbers.

[kukeltje]

PLEASE keep in mind that the webapp is NOT jBPM. It is a front-end for jBPM. It is not optimized to be used in high volume, high performance environments, rather for RAD purposes. So if you realy want to do a stresstest (loadtest?) of jBPM you should access jBPM on a lower level.

This does not mean that we are not curious for the results ;-)

Ronald

[jcaraneda]

Configuration environment:
1. server: one laptop with 1.6 GHz, app server JBoss 4.0.3, and 1 GB RAM.
2. simulation client: one laptop with 2.0 GHz, and 1 GB RAM.
3. manual test client: one laptop with 2.0 GHz, and 1 GB RAM.

Results: I can to simulate 340 users doing 1 click/second, i.e., 20400 clicks/second, by other side, sending request from the other client (called manual test client), the level service was of less than to 10 seconds/page.
I don't measured the band width used. With this level of requests the server processor near of 99% and RAM memory not increase more than a 3% from 0 request to 20400 request.

I wait this information could be util for you.

JC

[kukeltje]

340 user doing 1 click per second.... how does that result in 20400 clicks per second?

[kukeltje]

ok.... probably a typo on your side... 340/second is 20400 per minute

[jcaraneda]

thanks,
you have reason, is 20400 clicks/minute.

JC