Borland Software Corp. is out of the gate fast in early results from a new benchmark that tests the performance of Java-based software in a Web application.
The Scotts Valley, Calif., company recently completed testing of its Borland Enterprise Server (application server edition) Version 5.2 running with its JDataStore 6.03 database against the SPECjAppServer2001 benchmark. The software ran on a Dell Computer Corp. server equipped with two Xeon processors.
The published results indicate that the Borland configuration ran over 112 transactions per second for a 30-minute period. Borland officials pointed out that the system was running in a single-node configuration, which they said was cheaper to operate because it was simpler to set up and simpler to manage than a dual-node configuration.
The SPECjAppServer2001 benchmark measures the performance of J2EE application servers connected to a database server. The benchmark, released in September by the Standard Performance Evaluation Corp. (known as SPEC), was derived from an earlier J2EE performance benchmark called ECperf 1.1.
Specifically, the SPECjAppServer2001 benchmark tests the Enterprise JavaBeans container in a J2EE 1.2-compatible server and models an environment that includes manufacturing, supply chain management and order-entry tasks. The results are stated as BOPS (business operations per second), which is the number of customer order transactions plus the number of manufacturing work orders divided by the number of seconds measured. A higher BOPS score is better. The benchmark also states results as a price/performance ratio; in this case, a lower number is better.
Given that the benchmark has only been around for two months, SPEC has published results for only a handful of configurations.
The Borland single-node configuration achieved a throughput of 112.33 BOPS with a price/performance of $500.36 per BOPS. The only other single-node configuration listed on SPECs Web site is from IBM and measures the performance of that companys WebSphere application server running on eServer iSeries hardware. It clocked in at 804.09 BOPS and $4,847.59 per BOPS. Of the other five dual-node or multinode configurations listed on the Web site, four achieved greater BOPS (which is not surprising since they used more hardware), and three had a higher price per BOPS. A configuration from Oracle Corp. and one from Sybase Inc. both reported price/performance ratios of under $400 per BOPS.