[email protected] (Ted MacNEIL) writes: > I've been doing capacity planning since 1981. > VM is better than MVS, but it's not 100% accurate. > > No software monitor can be.
re: http://www.garlic.com/~lynn/2014b.html#78 CPU time it may not be reproducible because of things like cache effects ... which would affect hardware monitors also. its not a software monitor issue ... it is whether the kernel diligently does the clock operations for every piece of work. MVS (and vm370) gets total cpu busy by clocking in and out of wait state and subtracting it from elapsed. however vm370 does that also for every other thing it does also ... so all the accounted for time plus wait state time should come up to elapsed time (there may be tiny slop doing the clock instructions ... or if PR/SM underneath is doing something) the detailed MVS capture ratio discussions imply that MVS isn't even bothering to do the clock accounting for large parts of the kernel (and in some cases has been as high as 60%). -- virtualization experience starting Jan1968, online at home since Mar1970 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
