We have an 11050 ver5.00 build 45 acting as the front end for 4 windows 2000 servers running Weblogic 5.1 SP13 for PeopleSoft tools 8.20.
Our content rule has advanced-balance ssl and balance lconn configured.
We are load testing the server using a perl script that generates persistent SSL connections to the web servers.
When we hit a stand alone system we can handle 700 user test over a 60 sec period with the Java thread count rising and falling immediately based on load. If we place the one server behind the CSS we get the same results. If we put two servers into the content on the CSS we get a higher thread count and a longer hold time for the threads. Presumably the CSS is holding the connections open longer and therefore not releasing the thread on the weblogic server.
Any ideas as to why? In this case the CSS adds load to the servers instead of balancing it.
