Berin Loritsch wrote: > When I get the ContainerManager done and a new scratchpad > implementation of Cocoon (as extending AbstractContainer), Main, and > CocoonServlet; we can start to assess the real limitations that the > current component model has with scalibility. > > Until then, you are going to do alot of work for only marginal gains.
I made some findings that seem to support your thread contention analysis - during some tests of Cocoon with the help of JMeter, I found staircase-like patterns in the resulting graph presenting the response times, starting as low as 20 ms and going up to more than one second; I tested with 40 interleaving parallel threads accessing /cocoon/xscript/sample each 300+100ms on the following setup - Machine: P4 1.5 Ghz 512 MB RAM - OS: SuSE 7.2 Linux - Java: Sun, Java 1.4.0 HotSpot Client VM (build 1.4.0-b92) - Servlet-Container: Tomcat 4.0.1 BTW, initially, I used Blackdowns Java 1.3.1 (after Suns version seg'faulted at the beginning of building Cocoon via Ant) but a 8 hours stress test (~ 4.2 GB of Cocoon-generated web traffic) produced a VM crash due to a garbage collector bug during a Socket.accept operation; consulting Javasofts Bug Database, this seemed to be a well-known garbage collector related issue and was apparently resolved in Java Suns 1.4.0, so I switched; up till now I had no VM-related issue. Best regards, Michael Hartle, Hartle & Klug GbR --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, email: [EMAIL PROTECTED]
