Process and CPU information

[G V R Shankar]

Hi All,

Following command, prints the unique values of CPU numbers.

ps -ef|awk '{print $4}'|sort -r |uniq

C
3
20
2
12
1
0

In my system, I have CPU0, CPU1, CPU2, CPU4, CPU6 and CPU7.

I don't uderstand the CPU numbers 20,12,3 which are not present in my system.

Could you please explain about this?

Ravi

[rariasn]

Hi Ravi,

C isn't the number of cpu's.

%cpu %CPU cpu utilization of the process in "##.#" format.

Currently, it is the CPU time used divided by the time the process has been running (cputime/realtime ratio), expressed as a percentage. It will not add up to 100% unless you are lucky. (alias pcpu).


Rgs,

[Pete Randall]

A quick review of the man page indicates that this "C" column indicates the "processor utilization for scheduling". I think you've mis-interpreted.


Pete

[rariasn]

Yes,

pcpu

The percentage of CPU time used by this process during the last scheduling interval. The default heading for this column is %CPU.

Regards,

[Dennis Handly]

> ... | sort -r | uniq

No need to use uniq(1) here, you can use: ... | sort -r -u

[G V R Shankar]

Hi All,

I am sorry, I could have presented my question in more approprite way. Even it is my understanding that CPU cloumn would show the CPU number, instead of number of CPU's.

In my system, I have CPU0, CPU1, CPU2, CPU4, CPU6 and CPU7. In that case, ps -ef o/p shoudl have 0,1,2,4,6 and 7. Please correct me if I am wrong.

Apart from that, I see 20, 12 and 3 numbers in CPU column.

I have little knowledge about pcpu, but I am interested in the CPU cloumn.

Cheers,

Ravi.

[Pete Randall]

> it is my understanding that CPU cloumn would show the CPU number, instead of number of CPU's.

That is wrong. The CPU column shows processor utilization not processor number.

>Please correct me if I am wrong.

You are wrong (see above).


Pete

[Dennis Handly]

>Following command, prints the unique values of CPU numbers.

Are you thinking of the CPU number in the top(1) output?

[G V R Shankar]

Thank you for correcting me. I was in a impresession that it was the CPU number as shown top command.

Out of 6 CPU's, if 2 CPU's utilization is above 90% (as per OVPA reports), how do I know which processes are keeping those 2 CPU's busy.

Can you please help me?

Ravi.

[Dennis Handly · Accepted solution]

>if 2 CPU's utilization is above 90% (as per OVPA reports), how do I know which processes are keeping those 2 CPUs busy?

Doesn't top(1) tell you which process is busy and using which CPU? (Provided you aren't using threads.)
Any reason you care?

[G V R Shankar]

Hi Dennis,

you are correct, top gives me the data I am looking for. I almost forgot the handy tools that we have in unix. I have been using the following command to get top 20 CPU consuming processes.

UNIX95= ps -e -o "user,pcpu,cpu,vsz,pid,ppid,args" | sort -rnk2 | head -20

If I look at TIME in top man page it says teh following.

TIME Number of system and CPU seconds the process has consumed.

The time column keeps updating when the process state is in run state (makes sense). The moment it goes in to sleep state, the time cloumn doesn't update. The moment it comes in to run state, it doesn't reset the time to zero (until and unless it is new process). It just adds the time to the privious value.

How would I know, how much time the process has been in run state before it actually went in to sleep.

%CPU is used to sort the top process.

Also, could you please explain me the difference between %CPU and %WCPU. What is Raw and Weighted.

Cheers,
Ravi.

[James R. Ferguson]

Hi Ravi:

> How would I know, how much time the process has been in run state before it actually went in to sleep.

I don't think you can determine that very easily (at least not with 'top'). A far better tool for "drilling down" is 'glance'. In fact, since 'glance' is architected to understand the HP-UX kernel, it is probably the best tool for performance analysis on these systems.

> Also, could you please explain me the difference between %CPU and %WCPU. What is Raw and Weighted.

The "raw" value would be a value for that sampling interval only. A "weighted" value would take into account previous values to arrive at a "smoother" average for comparative purposes.

If you Google for "weighted average" you will find (among others):

http://en.wikipedia.org/wiki/Weighted_mean

At the least, this will give you a better understanding of "weighted".

Regards!

...JRF...