Using a Java benchmark to break bottlenecks

• By Jack Vaughan
• September 24, 2002

[PROGRAMMERS REPORT -Sept 24, 2002] - The Standard Performance Evaluation Corp. (SPEC) has released SPECjAppServer2001, a benchmark for measuring the performance of Java 2 Enterprise Edition (J2EE) application servers connected to databases. The benchmark enables users to test and compare hardware and Java software platforms, and to check the performance of their Java server applications.

The benchmark focuses on testing the EJB container in a J2EE 1.2 server, using a workload that models a supply-chain, order/inventory and manufacturing environment.

The SPEC benchmark is derived from the ECperf 1.1 (JSR-131) developed as part of the Java Community Process (JCP). The group, which has a long history in building and maintaining hardware benchmarks, has made some changes to the original Java server benchmark to keep it in line with its general benchmark policies.

SPEC's new stewardship of a Java standard should hearten those who distrust vendor-administered benchmarks. Of course, there is a body, which includes many developers, that looks askance at all benchmarks. Programmers Report asked Kaivalya Dixit, SPEC president, what such a benchmark means to a developer today.

'Application developers that use J2EE environments can benefit from this benchmark,' said Dixit. 'They can improve the performance of their application by analyzing performance bottlenecks and applying appropriate solutions to their software.'

Using a benchmark like this can help users to uncover architecture issues that may not usually be uncovered until an application is deep in the deployment process. Said Dixit: 'You can find out if your app server software is not fast enough, or if your database server's response time is too slow.

'A benchmark like SPECjAppServer provides 'a harness' to try different hardware and software configurations to optimize a solution,' he added.

As for the objectivity of the benchmark, Dixit is clear. 'SPEC's infrastructure of common code, comprehensive run and reporting rules, review process and tools provides a level playing field for the benchmark,' he said, noting that SPEC invites comment on all its benchmark methods.

SPECjAppServer2001 is priced at $2,000.

Links:
For more from SPEC, go to http://www.spec.org/osg/jAppServer/

For other Programmer Report articles, please go to http://www.adtmag.com/article.asp?id=6265