> -----Original Message----- > From: Caldarale, Charles R [mailto:[EMAIL PROTECTED] > Sent: Monday, February 07, 2005 10:04 PM > To: Tomcat Users List > Subject: RE: tomcat CPU utilization pegged...total CPU only 50% > [bcc][faked-from][bayes] > Importance: Low > > > From: Kevin Williams [mailto:[EMAIL PROTECTED] > > Subject: RE: tomcat CPU utilization pegged...total CPU only 50% > > > > The two Xeon CPU's are hyperthreaded to represent 4 CPU's > > on the machine. I'm not sure exactly how tomcat/java > > interact with this, but it was not favorable. > > Other than for determining the number of threads to use for parallel GC, > the JVM does not care how many CPUs there are. I don't know what Windows > returns for the number of CPUs installed when hyperthreading is enabled, > but you can determine that with the Runtime.availableProcessors() method. > > You really should be measuring performance by some measurement of actual > workload completed, not CPU utilization. Turning off hyperthreading may > actually reduce your overall throughput. (Then again, it might make it > better - very application dependent.) > >
You're right. I was measuring average response time. The CPU was just the bottleneck. My average response time under a specified user load actually decreased with hyperthreading off. My next step is to start analyzing the code:) Thanks for the pointer! Kevin Williams --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
