SUMMARY: vmstat and top interpretation, performance issues

From: beginer unix <unixbeginer_at_hotmail.com>
Date: Wed Nov 21 2001 - 22:48:19 EST
Thanks to:
Jay Lessert, Darren Dunham, Steve Sandau, Kevin Buterbaugh, Thomas Knox , 
Frans Tigelaar,  Daniel David Benson and Ric Anderson

I am enclosing the reply by Jay Lessert
>Hi,
>
>I am failing to interpret the following.. vmstat output consistently shows
>0% cpu idle time as shown below.
>
>     cpu
>us  sy  id
>67  33   0
>
>I did a top and found out that a single process is using up 99.6% of cpu
>time.
>
>But load average seems Ok. it is around 1.0 on a single CPU sparc machine.

That's right.  Load average is "the average number of jobs in the run
queue" over some period of time.  The load average output displayed by 
uptime(1)
and top is for 1, 5 and 15 minutes, I think.

So a system that is running one well-behaved cpu-bound process, with
nothing much else happening will have a load average of awfully close
to 1.0.

>Whatz the reason behind this... Some times on a different machine, I found
>that load average shoots upto 5.0 but my vmstat shows the following
>
>     cpu
>us  sy  id
>8   3   89

Ok, so you've got an average of 5 jobs in the run queue, but some other
system resource (waiting for I/O, waiting for paging) is keeping them
from being cpu-bound.

>Which one of above is more worrying as far as performance is concerned?
>High load average or cpu idle time being 0?

The 1.0/0% case is clearly not worrying, right?  You're running one
job, it's using all the cpu, that's great.

The 5.0/89% case *might* be worrying, although if for example, the five
jobs were spending all their time waiting for five tape drives, it
would be fine.  On the other hand, if the five jobs *should* be
cpu-bound, but they're out of RAM, then there's a problem to deal
with.


Sunil

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
Received on Thu Nov 22 03:48:19 2001

This archive was generated by hypermail 2.1.8 : Wed Mar 23 2016 - 16:32:36 EDT