When garbage collection touches all those pages with such frequency, I would imagine that z/VM would want to consider them not eligible for paging. Even when a WAS virtual server is inactive, the working set size never shrinks, as would be expected.
I've suspected this CPU and memory behavior in our WAS 4 and WAS 5 instances, and we'll investigate it further. It would appear that a direct port of an Intel architectured application may not play well with others in a virtual world. Perhaps software architects who are about to port to Linux on zSeries should spend some time in Hursley, England to see how the CICS folks learned to share. On Monday 07 June 2004 12:27, you wrote: > Anyone using WebSphere 5.1.0.3 on z/linux seeing an garbage > collection problem in the 1.4.1 JVM that ships with the product? > > We're seeing (and have a PMR open) where java garbage collection is > running at intervals of 100-1500 milliseconds to do cleanup on a 512 > MB heap size, even when that heap is 75-96% free. > > We turned verbose garbage collection on and discovered this. We're > spending 65% of the process time in garbage collection when we > shouldn't be. > > --------------------------------------------------------------------- >- 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 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
