Go Premium for a chance to win a PS4. Enter to Win

  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 601
  • Last Modified:

How can I accurately graph the average CPU usage... not just load average?

I am writing a perl script that does some forking work, and I would like to monitor the loads of the computer running it.  The load averages are typically around 3 - 6 over 1, 5, 15 minutes, but I seek more information than that.

I already watch the CPU usage on a 10 second interval, as presented by "ps -axo %cpu", but I don't think I am getting the whole picture.

When I use "top" it shows % User, % System, % Nice, % Interrupt, % Idle, but I don't think that "ps -axo %cpu" shows anything but the % User.  How can I get a more accurate picture of how much load the system is under?

I haven't been able to find a perl module that does this.

Thank you.
2 Solutions
You could try

time perlscript
There are multiple meanings for the phrase "how much load the system is under". Since it sounds like you're trying to get a more comprehensive picture rather than a single "number", you might like a utility like 'vmstat' or combine it with 'iostat'.

Within your application, you can use the 'times' operator to get information similar to what is reported by the command-line utility 'time'.
I have a suggestion for you.

Once you have chosen your metric, you could very easily write a perl script that call "top" periodically, pick up the relevant information and print it in a file. Say you chose, the "memory usage". You call top every second and pick up the "memory usage" of your script. Now print all this in a file.

After you are done with the script, you could use "gnuplot" to plot a graph of this data which gives you a broader picture or the trend of memory usage by your script.

Of course, once you have such a monitoring script in place, you could make it look for multiple parameters at a time.
What does it mean to be "Always On"?

Is your cloud always on? With an Always On cloud you won't have to worry about downtime for maintenance or software application code updates, ensuring that your bottom line isn't affected.

Hi PerlKing,
> call "top" periodically

You can simply open the output of top, like:

open(TOP,"top -b -s10 -d15 -u 0|") or die "No top";

And then just parse the output. An empty line ends a single output.

(check top man pages for the options used)

Sorry I was not clear with my wordings, but that is exactly what I meant too!
Thanks for clarifying it stefan.
try parsing the contents of /proc/stat yourself instead of using top.
the first line will give you counters for overall cpu usage.
A sample:

cpu  3028583 89870834 2870091 9335

Then use something like:
         ($cpu[$r], $usr[$r], $nice[$r], $sys[$r], $idle[$r])= split / +/,$line;

when you take a new measurement. flip $r
         $r = 1 - $r
and convert the last two measurements to a percentage:

      $usr_100  =rate_prcnt($usr [$r],$usr [1-$r],$total[$r],$total[1-$r]);
      $nice_100 =rate_prcnt($nice[$r],$nice[1-$r],$total[$r],$total[1-$r]);
      $sys_100  =rate_prcnt($sys [$r],$sys [1-$r],$total[$r],$total[1-$r]);
      $idle_100  =rate_prcnt($idle[$r],$idle[1-$r],$total[$r],$total[1-$r]);

These four numbers give you the average cpu usage during the time between the two measurements.

sub rate_prcnt {
   my ($a1,$a2,$b1,$b2) = @_;
   if (!($a1&&$b1&&$a2&&$b2)) {
      return 0;

   return (abs($b1-$b2)==0)?0:sprintf ("%.2f",100*abs($a1-$a2)/abs($b1-$b2));

Note that I grabbed the code for this from procallator.pl, which is a contribution to Orca, which is a systems monitoring and graphing package using RRD.

Featured Post

Concerto Cloud for Software Providers & ISVs

Can Concerto Cloud Services help you focus on evolving your application offerings, while delivering the best cloud experience to your customers? From DevOps to revenue models and customer support, the answer is yes!

Learn how Concerto can help you.

Tackle projects and never again get stuck behind a technical roadblock.
Join Now