On Mon, Apr 25, 2011 at 8:44 PM, Ron Foster at Baldor-IS <[email protected]> wrote:
> We are a SAP shop and do not run WebSphere. However we do run some SAP > Java instances. On some of those systems, we have found it beneficial > to define two CPUS, and let the garbage collector know about it. > Depending on the size of your heap, sometimes the the garbage collection > times improve noticeably. One of the problems with Java Garbage Collection is that it halts other work in the JVM for the duration of the GC (because you're moving stuff around, you can't have people trying to use it). While GC has improved over time to reduce the amount of objects handled during each scan, my understanding is that it still suspends other JVM work. Part of GC can exploit additional CPUs when you activate the parallel GC (not the default). If you have limited CPU resources available on z/VM, spreading your CPU resources over 2 virtual CPU means a single thread runs slower than on a 1-way guest. So unless you have excess CPU resources, virtual 2-way would not make it faster and possibly slower. If you find it improves performance anyway, then there may be different factors involved. And that's what drives the virtual-MP discussion in general. You only give the guest more virtual CPUs when - the workload can take advantage of it - and the business justifies it - and when you expect the resources to be available when the guest wants it Your performance monitor shows you what was holding the guest back in times where performance was critical. Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
