Subject: Re: vmstat output is wrong
To: Brian Stark <bstark@siemens-psc.com>
From: David Brownlee <abs@anim.dreamworks.com>
List: current-users
Date: 04/22/1998 12:02:36
	Given that the 'cs' column is part of the 'faults' section, my guess
	is that leaves us with rounding errors.

	In the case given, I've listed the total, and a _possible_ set of
	real values that might have resulted in the numbers shown

		> us sy id	(total)	   (real  us  sy  id )
		> 64  2 34	100
		>  1  0 99	100
		>  2  2 96	100
		>  5  2 94	101	         4.6 1.7 93.7
		> 12  7 80	 99		12.3 7.3 80.4
		>  0  0 99	 99		 0.3 0.3 99.4

	Of course.. I could be totally off base here :)
	If this is the case, and AIX 'fudges' the numbers to make them add
	up to 100%, NetBSD is actually being 'more accurate'....

		David/absolute

	  -=-  Just adding to the wrinkles on his deathly frown  -=-

On Wed, 22 Apr 1998, Brian Stark wrote:

> Hello,
> 
> I have noticed what appears to be a bug in the output from the vmstat
> program and would like to know if people running -current have the same
> problem and also if this is present on the other platforms.
> 
> First, according to the man page for vmtat, part of the information that
> vmstat shows is a "Breakdown of percentage usage of CPU time." However,
> when I run vmstat on NetBSD 1.3/i386, I see output like this under the
> cpu section:
> 
>  cpu
>  cs us sy id
>  57 64  2 34
>  87  1  0 99
>  75  2  2 96
>  77  5  2 94
> 121 12  7 80
>  28  0  0 99
> 
> Why does each row under the cpu section not add up to 100%? When I run
> vmstat on the AIX workstations at my office the values under the cpu
> section do add up to 100%. However, I did notice that vmstat on a Sun
> workstation returns strange values too:
> 
>       cpu
>   cs us sy id
>  270 19  8 73
> 1098 10 14 76
> 1115  3 16 81
>  791  9 13 78
> 1334  6 26 68
> 
> 
> What makes NetBSD different??