On Tuesday, 01/27/2009 at 01:43 EST, "Schuh, Richard" <[email protected]> wrote:
> We have some questions about MIPS consumed by VM and Linux from management. > They want numbers based on the cost of the MP effect by operating system, > expecting different numbers for VM than for Linux (I haven't yet said anything > about considering real MP effect for VM and virtual MP effect for guests having > more than 1 virtual cpu). Is there any acceptable way to convert CPU Model and > number of engines into MIPS? For example, our main VM system, one that has no > Linux guests, runs in a 7-cpu LPAR on a 2084-507, while there are two LPARS on > the same CEC, sharing 3 IFLs, running Linux under VM. Does this help? (It's z/VM 5.3) http://www.vm.ibm.com/perf/reports/zvm/html/mt31cpus.html I infer from the chart that up to 6 CPUs, the MP effect is nearly unnoticeable. Of course, YMMV, depending on workload and machine type. Alan Altmark z/VM Development IBM Endicott
