The data provided by the z/VM Monitor is VERY correct. It's only the interpretation that may need "tuning".
The issue with Linux is that Linux does not use the s/390 facilities needed for accurate accounting when running in a virtual environment. In a VM environment, the capture ratio should be above 99.9%. This is true with VM running native or in an LPAR. For storage, capture ratio should be above 95%. Capture ratios are very often very good validaters for performance monitors. A very high capture ratio provides a very good confidence factor - you know what resources is used, and by what.... > >From: Paul Kaufman <[EMAIL PROTECTED]> >Are the measurements that z/VM provides on guest machine (Linux server) >resource usage also incorrect? > >On Monday 22 July 2002 14:59:48 PDT, Barton Robinson wrote: >> There are some issues with monitoring linux under VM - the >> numbers are not quite right, well, they can be absolutely >> wrong. >> >> Linux does accounting and CPU measurements for processes, etc as >> if Linux owns 100% of the processor resource. Thus all of your >> Linux CPU numbers will be inflated. My example that I've >> presented at SHARE, IBM Tech conferences and assorted other >> meetings, shows Linux off by 1 order of magnitude. (This >> includes TOP, other agents, any way you like to measure Linux >> CPU) If you would like valid linux performance numbers when >> running on VM, the only solution on the market (i'm pretty darn >> sure) is ESALPS, from Velocity Software (long time vendor of VM >> performance products....), and for the last 14 years, Velocity >> has ALWAYS supported new releases of VM. > ><snip> "If you can't measure it, I'm Just NOT interested!"(tm) /************************************************************/ Barton Robinson - CBW Internet: [EMAIL PROTECTED] Velocity Software, Inc Mailing Address: 196-D Castro Street P.O. Box 390640 Mountain View, CA 94041 Mountain View, CA 94039-0640 VM Performance Hotline: 650-964-8867 Fax: 650-964-9012 Web Page: WWW.VELOCITY-SOFTWARE.COM /************************************************************/
