On Wed, Jun 24, 2009 at 12:18 AM, Spracklen, Ken<ksprack...@mt.gov> wrote:
> I would appreciate any words of wisdom on ways to cut the batch processing > time on the zLinux guest to be comparable to the blade server. For starters, you can collect the per-process CPU statistics and identify the most heavy threads in a javacore dump. That's mostly off-line work and does not put additional load on the system like profiling tools would do (if they even provide you the correct data in this environment) One of the areas I worked on is the implications of virtualization on the IBM JDK profiler.There are some real issues there (no pun intended) but they tend to show during application startup. Since you have entirely different installations to compare, there could be some extra unrelated stuff that got installed and takes resources. I would not be worried by z/VM doing a few pages per second, but it is important to know whether your server is in page wait. Most likely your idle servers don't drop from queue and frustrate z/VM memory management. As long as you have sufficient expanded storage that may still be acceptable. Lot of questions, and we'd probably need full z/VM and Linux performance data to understand what's happening. Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
