Joined: 17 Mar 2004 Posts: 8 Topics: 5 Location: Madrid-Spain
Posted: Fri May 28, 2004 5:44 am Post subject: CPU usage over 100%
Hello folks,
We have a 2066 with only 1 proccessor shared by three z/OS 1.4 systems. On getting a summary report with RMF for two of them, we are obtaining some 15-minutes intervals in which, adding the values of CPU busy field for this interval, it results in something over the 100% (we have obtained even a 130 % in some intervals). Tell me if I'm wrong on thinking this is the average value for this interval, not a peak value. Can anyone explain me why it occurs? I don't know if it has something to do with the algorithm used to calculate it, of if the "CPU busy" field considers in some way the requested CPU instead of the actually used CPU.
Joined: 01 Jun 2004 Posts: 3 Topics: 0 Location: Melbourne
Posted: Tue Jun 01, 2004 4:12 pm Post subject:
Alberto,
I'm quite new to mainframes, but I've heard about this somewhere.
You're using MXG and SAS to make these reports I assume? The problem is in MXG apparently, something is not quite right with the calculations...
I'm not sure of the details, as it was someone much senior to me talking about it, but I got the impression that there was a known problem here.
Also, (exuse me if this is an amatuer question!), what is the actual metric that is returned on CPU usage by RMF? MIPS, CPU %, CPU minutes?
I'm just thinking that perhaps, if you are running multiple systems, the CPU weight on a certain LPAR is considered to be "100%" and the metric returned is a percentage of that. So if you had 100% utilisation on all three of the LPARS, and you added them all up, you'd get 300%.
Joined: 01 Jun 2004 Posts: 3 Topics: 0 Location: Melbourne
Posted: Wed Jun 02, 2004 8:32 am Post subject:
From the page supplied by semigeezer:
Quote:
4. If the CPU utilization is greater than 100%, then the number above 100 indicates the number of ready address spaces not receiving service. For example, a RCCCPUT value of 107 indicates there are 7 address spaces in the domain that currently are waiting to receive service. If the value is 128, then there are at least 28 address spaces waiting to receive service.
This is only a snippet of much more information, and I hope I'm not taking it out of context, but it would appear to be your answer, Alberto...
One thought - if a single address space correlates to 1% of CPU, what happens when a CPU doubles in speed? Obviously an address space doesn't then take 0.5% CPU? I guess I need to go and read that page properly....
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
FAQ
Search
Quick Manuals
Register
Profile
Log in to check your private messages
Log in
