Sorry for the repetition but I pressed submit before I finished :P
I'm running a custom benchmark which simulates an intensive database access to a J2EE aplication I've migrated from OpenEJB 1.4 + Tomcat 5.0 to JBossAS 4.0.4:
A single client, requesting 1, 10, 100, 1000 and 10000 ocurrences of the test case;
Multiple clients (1, 10 and 100 threads) requesting 1, 10 and 100 ocurrences of the test case (to test the connection pool).
The results I've got so far are very disappointing. JBoss is up to 2 x slower than OpenEJB + Tomcat 5.0 (in my test, of course).
The J2EE aplication uses the DAO pattern for persistence.
The DBMS is MS SQL Server 2000 and I'm using the jTDS driver.
The test case simulates a set typical uses for the aplication (only reads).
I've made some optimizations (log4j level = ERROR & file only) but it doesn't seem to have a significant impact on the performance.
Any suggestions / questions / help would be apreciated.