Subject: Re: Data collected about detailed usage
From: Tom Duerbusch (DuerbuschT@stlouiscity.com)
Date: Wed Nov 04 2009 - 10:49:18 EST
The point was, that nothing has changed, in some products. When you sample the CPU, it works great with a single image. But you have no idea of what is happening in other images. Hence CPU seconds varies wildly with the same job. IESX has real numbers. But at the end of step, the performance numbers are initialized. IMHO, that is why DA shows an *, as IBM couldn't report a reliable number. The rate of CPU sampling should always be based on the speed of the processor. Tom Duerbusch THD Consulting >>> Rich Smrcina <rsmrcina@wi.rr.com> 11/3/2009 8:19 PM >>> So you're comparing a discussion you had about this over 20 years ago with the current state of the software? And as to CPU sampling... how quick is quick? Quick isn't a measurement. What was quick 20 years ago is glacial now. Tom Duerbusch wrote: > At least with CA-FAQS, the CPU time shown isn't real CPU time used. I seem to recall this discussion back in the '80s. The easiest way of determining CPU utilization was to sample the CPU. If it came back quick, the CPU was idle. If it didn't come back quick, it was busy. With one active machine, under VM, this worked fine. > > However, when I have multiple VSE machines very active (running batch), each, let's say, getting 25% of the processor, the FAQS D S command shows near 100% utilization on each of the machines. And then assigns to each of the active partitions within its own machine, CPU figures to match the system utilization. > > This also happens with PCS. When you do a "z" on an event. The exact same job may run from 80 CPU seconds to 120 CPU seconds. > > The D A in the II panels, do take in account the effect of other machines, or, at least, pick up the "near real" numbers from VSE Job Accounting. > > More recently, this was a problem (and solved) with Linux under VM. > > Tom Duerbusch > THD Consulting > > -- Rich Smrcina Phone: 414-491-6001 http://www.linkedin.com/in/richsmrcina Catch the WAVV! http://www.wavv.org WAVV 2010 - Apr 9-13, 2010 Covington, KY
This archive was generated by hypermail 2b25 : Sun Nov 22 2009 - 01:05:05 EST