On Wed, November 26, 2003 1at 1:17 am, Jeff Tulley wrote: > I had not thought of that, but that should work well. I'll try it out > and see. > > Right now I'm evaluating whether my test is actually a reasonable > real-world test. The reason our testers did not find this earlier, is > that their tests are a little more realistic to typical web site use. > My test was more of a top-performance benchmark of dynamic content only, > with each connection creating a new session. I guess my test is like a > JSP site being slashdotted. It would be good if the PersistentManager > solves the problem. > > Thanks for the good idea.
Glad I could help. While the PersistentManager should solve your test case and problem of being swamped by new sessions, I feel that a better performing solution would involve a Manager which allowed a set number of sessions to be kept in memory (or even better, able to watch the amount of memory used by sessions) and when a preconfigured limit is reached, swap out the LRU session to disk. At some point it may also be desirable to expire a session early, too. Although unlsss you are being slashdotted or store a lot of data in your sessions, its easier to buy another stick of RAM for the server and increase Tomcat's -Xmx setting for most users. -Dave --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
