Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 22 Aug 2005 16:36:11 +0200
From:      "pobox@verysmall.org" <pobox@verysmall.org>
To:        Mario Hoerich <spambox@MHoerich.de>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: top -I and load average from 4.9 into 5.4
Message-ID:  <4309E2DB.1070000@verysmall.org>
In-Reply-To: <20050822142101.GB74593@Pandora.MHoerich.de>
References:  <4309C9F7.3070904@verysmall.org> <20050822142101.GB74593@Pandora.MHoerich.de>

next in thread | previous in thread | raw e-mail | index | archive | help
>> We upgraded from 4.9 to 5.4 and the 'top -I' and load average seem to 
>> behave slightly different. 'top -I' in 5.4 stays empty most of the time 
>> with processes flashing in and out from time to time. 
> 
> Well, I'd actually expect them to, as -I filters out any idle 
> process.  Justed tested on my 6.0B2, major CPU hogs are shown
> as expected.

That's clear. Just on the same server under 4.9 somehow the processes 
tend to 'stay longer' and not just flash for parts of the second. Now 
under 5.4 the list 'top -I' is most of the time empty while the CPU goes 
from 0% to 20% and back all the time. And just very seldom a process 
flashes in the list. 'top' under 5.4 shows all the time 3-5 active 
processes at the top of the list.

>> In 4.9 the processes stayed there until they are done.
> 
> Where's the difference to top [-i], then?

'top' shows a long list with active and not active. 'top -I' shows just 
the active ones, but as some take few seconds - they do stay there for 
few seconds (under 4.9). Under 5.4 they either do not show or if they 
show - it is for parts of the second.

> I don't really know enough about how load and utilization
> are calculated.  I'd guess load is calculated by the scheduler,
> based on the number of processes whining for more/larger time
> slices, whereas utilization is guessimated elsewhere, probably
> something like process CPU time / available CPU time.  
> 
> I'm none too sure there's an actual correlation between these
> values, let alone load 1 = utilization 100%.

I had this feeling. So that's clear. Just the question is now why 'top 
-I' under 5.4 is so quiet. Has something changed in 5.x or is it a 
question of setup?

Iv




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4309E2DB.1070000>