Performance Benchmarks

KeyLabs recently benchmarked the Java System Web Server 6.0 and 6.1 against Apache 2.0, Tomcat, and OpenSSL. The ratio of static to dynamic (e.g., JSPs) content was 70/30. The test results are summarized below. The full report is available online.

Metric: Concurrent Users per Server
Rationale: This metric measures connection handling efficiency. The more concurrent users per server, the better the return-on-investment on the hardware and software.
Results: The Java System Web Server can handle 50% more concurrent users per server, 200% more when using SSL. In effect, the Java System Web Server requires 50% less hardware to handle an identical workload as compared to Apache and Tomcat.

Metric: Page Load Times
Rationale: This metric measures raw performance. Defined, it measures the time it takes to load all elements of a page, and is an aggregation of connect, send, response, and process times.
Results: The Java System Web Server is 8x faster, 24x faster when using SSL.

Metric: Error Rates
Rationale: This metric measures error rates (e.g., 404 errors) under heavy load. Supporting peak/flash crowds are critical, especially in e-commerce or customer service environments.
Results: On average, the Java System Web Server had 24x fewer errors than Apache and Tomcat. KeyLab notes, "The Apache/ Tomcat configuration did not respond as well to rapidly increasing load level."