On Sat, Oct 31, 2009 at 1:33 AM, Martin, Terry R. (CMS/CTR) (CTR) <terry.mar...@cms.hhs.gov> wrote:
> Thanks Ed. I am not running SFS and I understand that VMRM requires the > config file to be under SFS control is this correct? Also the only thing > that came on the A disk for the VMRMSVM user was the PROFILE EXEC. I read > that there should be a sample config file as well as some other files on the > A disk also, is there another place theses file can be found? The MONDCSS size really does not matter. It must be configured large enough to hold full sample configuration data so that you can make sense of the sample data. When the MONDCSS is large enough, the default partitioning will leave enough data for sample configuration. Otherwise you must specify the startup parameters. Once that is done, it should be good enough for all. When the amount of monitor event records is high (seeks enabled, or many virtual machines dropping from queue often) it may be a challenge for a virtual machine to consume the event buffers before they expire. With multiple virtual machines processing the same monitor data, you have more work to be done in a timely manner. If you can't keep up with event buffers, the bottleneck is normally the number of outstanding IUCV msgs rather than the available pages in MONDCSS. This is addressed with the parameters for the monitor rather than the size of the DCSS. If I were looking at VMRM, I would also investigate what happens when the system is so constrained that VMRM does not get the resources to process the sample in time (that happens sometimes when CP has paged out the MONDCSS). It would be interesting to see if VMRM at that time can undo the settings it did before that got z/VM in the thrashing situation... Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/