sar -q output

[Ninad_1]

Hi Gurus,

Please can anyone explain me the terms in the sar -q output
i.e. runq-sz , %runocc , swpq-sz , %swpocc
I am not able to understand the meaning of the metrics and their use for interpretation of performance in the real world.
Does runq-sz means processes currently running on the system ?
and swpq-sz mean processes waiting to be run but are ready to run as soon as cpu can be allocated ?
and what are the occ metrics ?
How one can use these to interpret something.

Attached is the output from sar for a particular day. Please can anyone explain me this in the context of the output attached.

Thanks
Nad

[marie-noelle jeanson_1]

Nad,

Have you looked at the man page:
http://docs.hp.com/en/B2355-90692/sar.1M.html

runq-sz Average length of the run queue(s) of processes (in memory and runnable);

%runocc The percentage of time the run queue(s) were occupied by processes (in memory and runnable);

swpq-sz Average length of the swap queue of runnable processes (processes swapped out but ready to run);

%swpocc The percentage of time the swap queue of runnable processes (processes swapped out but ready to run) was occupied.

Marie.

[Ninad_1]

Marie,

Yes I agree thats the first place anyone should look for and I have read the man page - but I have not understood the meaning and how to interpret the output I have attached to determine the performance of the system.
Any help is appreciated.

Regards,
Nad

[Chan 007]

Ninad,

Give me your OS version, memory/swap and number os CPUs

to give complete information.

runq-sz - if the output parameter is more than 2, then the CPU is loaded.

%runocc - if this output is more than 90 some say, but I recon more than 75% then the CPU is loaded and will degrade the performance. This does happen if any zombie or a breakaway process/looping/errorous will tend to do this activity. you can see this process in TOP/glance. Either kill it. Or you have to reboot the system taking an outage.

swpq-sz - if is the swap queue, this should be same as CPU

%swpocc - this persent same 75% and above is your system uses more swap. you have to add more to it.

Hope this helps...

Chan

[Ninad_1]

Hi Chan,
Sorry for my late update
OS - HP-UX 11.00
V2500 server with 18 * 440MHz CPU , 24 GB RAM
Swap - 16 GB on device and 18.8 GB pseudo swap.

I disagree with you on one point -
You said "%swpocc - this persent same 75% and above is your system uses more swap. you have to add more to it. " Does swpocc really show swap usage and can we determine if swap space is required. because I thought we can check through swapinfo and check vmstat to see if pi , po is there to see if paging is there and this will really occur if server falls short of memory. Please correct me if I am wrong.

Also what is your opinion on saq -q output of my system. is it OK ?

Thansk
Nad

[Chan 007]

Nad,

Sorry I did a cup/paste, What I meant was on the system swap (I am not talking about swapinfo), this is the processor swap that happens internally by the processor. Hope now you got me. (Swaping of each process that runs in one processor), This is related to time sharing or time slizing how ever you can call.

On your system report it is normal. You try using sar -M -q, which will provide you with each processor load too.

Chan

[Ninad_1]

Chan,

Alright.
Why I asked the question if my sar -q output seems fine is because in the vmstat output I have figures between 100 - 800 during a particular period of time in the r (processes in run queue) field , so I was imagining if this figure is same as that shown by sar -q ?
Please can you tell if the r field in vmstat and the runq-sz in sar -q are same ? What could be the reason that there are so many processes in run queue - are they running at the same time , why its not showing anything in the swap queue as all the processes cannot run at the same time and they need to be swapped in and out according to their allocated time slices.
Please advice.

Thanks
Nad