CPU States

Help
2004-04-20
2004-08-11
  • George Mellen

    George Mellen - 2004-04-20

    I have a fairly busy FTP server running on a 4 proccessor Sunfire 480.  I have  a question regarding the line labled CPU states. Specifically the idle percentage. Is that calculated based on a single CPU or the total? I am seeing single digit idle cpu pertentages when I may have only a handfull of FTPD sessions running at that time...... I find this number hard to believe.

     
    • William LeFebvre

      It is supposed to be calculated across all available cpus.  That is, the percentages when added will never exceed 100 (except for roundoff error) and 100% reflects the total amount of cpu power available on the system. This same holds true for the cpu percentage column in the process display.  In your case, a single-threaded process would never exceed 25% cpu utilization.

      You can verify (or refute) top's numbers by comparing them to the output of vmstat or iostat (the columns labeled us, sy, wt, and id).

       
    • Trick Hartman

      Trick Hartman - 2004-06-04

      When I run top on my Linux ppc64 machine, I receive the following default CPU states

      Cpu(s): 19.4% us,  4.7% sy,  0.1% ni, 73.6% id,  1.7% wa,  0.1% hi,  0.5% si

      What do the last 3 columns mean? wa, hi and si?

      I see that the states are described in other areas as "user" "system", etc.

      And I have read the man pages, but these values are not described there. (or so I think I have seen many different man pages. :)

       
    • William LeFebvre

      The top that comes with linux is not Unix top. It is something written specifically for linux and not written by me. Sorry but I cannot shed any light on your question about wa, hi, and si. Perusing the source would be your best bet. My guess for wa would be wait, possibly io wait. But for the others I have no guesses.

       

Log in to post a comment.

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks