today I took the day off, but tomorrow I plan on running more benchmarks. I ran into one weird behavior when I ran the same jmeter test with 10 threads for 5K iterations. Half way through tomcat stopped accepting connections, but was fine. If I wait two minutes, it continue to accept requests. That tells me tomcat is dramatically improved in stability. In the early days of servlet containers, I used to run stress tests until the servers crashed. Back in 99 most of the servlet containers including websphere, jrun and tomcat would crash a couple minutes after it stopped accepting connections. Now tomcat correctly denies connections and performance degrades gracefully. peter Remy Maucherat <[EMAIL PROTECTED]> wrote:Peter Lin wrote: > I forgot to mention the Jmeter settings I used. > All tests were with 1 thread for 5000 iterations. The first two image tests used >images for tomcat. The third image test with 194.5k image was a simple screen capture >of my desktop.
Thanks for the small benchmark. It is a bit odd that some tests are slightly worse, although it's well within the margin of error. If you have other tests (maybe with a higher load factor), send them. In 4.1.17, I'll provide a slightly more optimized default configuration for the connector (using disableUploadTimeout="true", and a bigger acceptCount, which may help avoid getting accept errors when under load). Remy -- To unsubscribe, e-mail: For additional commands, e-mail: --------------------------------- Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now
