Reading Glance/iX correctly (289 Views)
Reply
Frequent Advisor
Michael Simone
Posts: 52
Registered: ‎10-19-2000
Message 1 of 5 (289 Views)

Reading Glance/iX correctly

I'm somewhat new to MPE, and am trying to get a handle on using Glance to monitor system hogs. I have an N4000 with 4 400-MHz CPUs. When I see CPU spike near 100% on Glance, I pay attention (and try to keep my boss from panicking). However, I'm wondering about the CPU column on the Global screen--our "big" jobs often show 99% in this column, which according to Help represents the CPU used by that process in the last interval. But if I have several jobs showing 99%, what does that mean? Does it actually go to 400% because there are 4 CPUs (Unix does this)? Or is it not a good gauge.

I realize we can't live our lives watching Glance, but I need to give my staff a tool to watch for runaway jobs, excessive processes run by users, etc.

Thanks!
Honored Contributor
Eric Sorensen_1
Posts: 573
Registered: ‎10-08-2000
Message 2 of 5 (289 Views)

Re: Reading Glance/iX correctly

Hello Michael,

A process that takes all of the CPU is not necessarily a problem-- if no other process is asking for CPU, then the system will allow one process to use all of the available resources.

If there are active sessions (processes) in the CQ and a batch job in the DQ, then the system will first attend to the CQ processes, and if there is any CPU-time left, the lower queues' processes may get some CPU time. If there are no processes above it that ask for CPU, then even a batch job in the lowest queue can take 100% of the CPU (if the job has enough work for the CPU to do...). "System hogs" must be defined as a process in the same or higher-priority queue taking lots of CPU-time when another process is waiting for CPU. Often this can be resolved quickly-- just altproc that process to a lower priority...

A system with 4 processors can show a total CPU usage percentage (in the detail lines on the "Global" screen) of up to 400%.
A problem well defined is half solved.
Esteemed Contributor
Dan Clifford
Posts: 201
Registered: ‎03-25-1998
Message 3 of 5 (289 Views)

Re: Reading Glance/iX correctly

Michael,

Additionally, if you have only one job running and it is taking 99% of one of the CPU's that is all you are going to get. A single process cannot span multiple CPU's so they will remain idle. So, you will never see one process with 200% CPU usage for example.

Dan Clifford
Trusted Contributor
Lars Appel
Posts: 131
Registered: ‎12-27-1996
Message 4 of 5 (289 Views)

Re: Reading Glance/iX correctly

You also might want to make sure that you have the latest Glance patch installed. There was one version that showed bogus 100% values on something like every other screen refresh. When this happened, it would also show a (bogus) JSMAIN process with 100% CPU in the global screen's process list.

Patch GLALXX8 or later fixed this.
see www.google.com/search?q=lars+appel
Honored Contributor
Chuck Ciesinski
Posts: 830
Registered: ‎10-03-2000
Message 5 of 5 (289 Views)

Re: Reading Glance/iX correctly

Michael,

Without the patch mentioned, you are probably seeing the utilization of just one CPU. My guess is that you'll see numerous processes and they add up to way over 100% in the
cpu column. This was pointed
out last year in my paper and presentation at HPWorld as one
of Glance/iX weaknesses. If you'd like a copy of both, email me at the address below.

For this particular screen, individual cpu utilization, I
prefer SOS from Lund Performance Solutions. Lund also has a similiar product for HP-UX.

My $.04 for inflation

Chuck Ciesinski
cciesinski@hns.co
"Show me the $$$$$"
The opinions expressed above are the personal opinions of the authors, not of HP. By using this site, you accept the Terms of Use and Rules of Participation.