[mrtg] Re: Great idea for MRTG

[Shaun Amy, CSIRO TIP/ATNF]

Wouldn't it be the case though that if you split your MRTG configuration into
multiple configuration files that all get started at the same time by "cron"
then you would effectively have multiple rateup processes running and the
standard Unix scheduling would use multiple processors if the host had them.

I only run MRTG on a uni-processor host but I do run four difference
invocations of MRTG which means four rateups.

Shaun.

>	I was wondering how MRTG would run on a dual processor ( or even more )
> 
>Linux machine. And for my surprise, it seems to run the same way on a 
>single processor machine.
>
>	Well, I know the 'hardest' part of MRTG is to run rateup for all target
>s, 
>because it's very CPU intensive AND disk intensive too. The fact is that 
>nowadays disks are fast enough to feed more than one rateup at the same 
>time, and multi processor machine are not very rare on servers environment.
>
>	So, I was thinking on a new directive to MRTG on mrtg.cfg file, somethi
>ng like
>
>processors: X
>
>	So, when MRTG gathers the values, and start calling rateups, it would 
>launch X rateup processes at the same time, because the machine has X 
>processors, and kernel would get one rateup to each processor. This would 
>allow people to run more and more Targets on a single machine.
>
>	Well, what do you think ? It this possible ? Is this interesting ?

--
* To unsubscribe from the mrtg mailing list, send a message with the
  subject: unsubscribe to mrtg-request at list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg