WebSphere Application Server release to release comparison

The purpose of this test was to determine which level of WebSphere® Application Server would give us the best throughput performance. We wanted to determine which code level (6.1.0.0 or 6.1.0.11) would be the best to use for the rest of our tests.

Test case description

In this test we ran the Trade 6 (Trade) workload in our standard 3 tier environment using RHEL 4.5 for WebSphere and DB2® for Linux®. The difference between the two runs was the that one run used WebSphere Application Server version 6.1.0.0 and the other run used WebSphere Application Server version 6.1.0.11.

Table 1. Release to release comparison test cases

	WebSphere Application Server			Trade Results				DB2 Database
Distro	Code Level	#CPU	%CPU	#Users	Resp	%ETR	%ITR	#CPU	%CPU
RHEL 4.5	6.1.0.0	4	99.36	60	34 ms	100.00	100.65	4	50.03
RHEL 4.5	6.1.0.11	4	99.46	60	35 ms	99.91	100.46	4	48.43

Observations

Both levels of code are very close in throughput characteristics. WebSphere Application Server version 6.1.0.0 consumes the same amount of WebSphere Application Server CPU resource as level 6.1.0.11. We show the ETR value for code level 6.1.0.0 as 100%, the other ETR and ITR values are shown as their relation to this value. The values for ITR, ETR, are nearly identical, with less than 1% difference between the two code levels. For this study we did not provide analysis for results with less than a 2% difference because the results might have been a run to run variance rather than a performance difference.

Conclusion

From a performance perspective, WebSphere Application Server version 6.1.0.11 behaves the same as version 6.1.0.0. The newer code had no impact on the performance characteristics of our workload. We always want to run with the latest code release when possible, so all further runs were performed using WebSphere Application Server version 6.1.0.11, the newer code release.