I am also facing the similar issue.
Below is sample statistics for a load testing i executed.
| Milli Second | Number of Request | Percentage with respect to total request |
| <=1 | 9215 | 8.99 |
| >1 and <=2 | 30016 | 29.28 |
| >2 and <=5 | 41046 | 40.04 |
| >5 and <=10 | 18899 | 18.44 |
| >10 | 3324 | 3.24 |
| | | |
| Total Request | 102500 | |
I have also created a client program which makes RMI call to my JBoss application. I have commented my business logic in Session Bean and just giving success response directly from Session Bean.
From client, i am sending a packet with event timestamp value and in session bean i am comparing the timestamp received and current timestamp.
Through above I am calculating the time taken in milli seconds for request to reach my session bean. And same I am printting in log. (only one log per request)
So above statistics clearly shows that more than 60% of requests taking more than 2 milli second, which is not acceptable for my kind of application.
Please guide me as well for the solution.