[mythtv-users] High idle load average on revo 1610
Greg Cope
gregcope at gmail.com
Sat Jun 5 08:35:10 UTC 2010
On 5 June 2010 09:22, Christopher Kerr <mythtv at theseekerr.com> wrote:
> Because load average doesn't measure what most people think it does, probably.
Agreed.
> http://luv.asn.au/overheads/NJG_LUV_2002/luvSlides.html
>
> 1.10 isn't -terribly- high. Yes, on a single core machine, it means
> something is having to wait, but it may not be anything terribly
> important. If you sit back and watch for half an hour, what does the
> 15 minute load average say?
If it is hyperthreading, then it will appear as 2 CPUs and hence a
load of less than 2 would imply little real waiting for CPU.
A good rubric is that a load average that is less than the number of
CPUs (reported) is not really an issue.
CPU stats from the left side of vmstat (usr/sys/idle) are a better
measure of CPU activity, and there are other metrics for IO/swapping
etc..
Greg
More information about the mythtv-users
mailing list