[mrtg] MRTG/RRD Performance tests: preliminary results

Steve Shipway s.shipway at auckland.ac.nz
Tue Sep 14 00:09:06 CEST 2010


> 	I wonder whether the knee in the load/impact curve isn't
> 	beyond the range of your test setup.

Quite possibly;  I'm only running the test for 15min windows (although there is an initial unmetered cycle so that files can be opened and RRD files synched).

> 	I noticed a significant change in disk IO (at least as
> 	shown by CPU IO-wait figures) on migration from native to
> 	RRD mode.  The move was made under some pressure, as disk

I was expecting a much better improvement for RRD vs. Native; I think the RRD files generated nowadays are larger than those generated back with MRTG v2.10 (800dp vs 400dp?) which will increase IO slightly.  Also, the disk caching is probably messing things up a little.

> 	size in half a year, so today's figures of 785 devices
> 	and 10756 targets give an idea of the scale.

You're about the same size as us -- we currently have 711 devices, and 11427 Targets, although I've spread this over 3 nodes in a cluster (using Routers2 in clustered mode)

Steve

Steve Shipway
ITS Unix Services Design Lead
University of Auckland
Floor 2, 58 Symonds Street
09 3737599 ext 86487
 Please consider the environment before printing this e-mail 

 
> 	I wonder whether the knee in the load/impact curve isn't
> 	beyond the range of your test setup.
> 
> 	I noticed a significant change in disk IO (at least as
> 	shown by CPU IO-wait figures) on migration from native to
> 	RRD mode.  The move was made under some pressure, as disk
> 	interference was causing trouble to the nightly backup run.
> 	Consequently, recording the number of targets and devices was
> 	not a high priority.  That said, our network hasn't doubled in
> 	size in half a year, so today's figures of 785 devices
> 	and 10756 targets give an idea of the scale.
> 
> 	Best regards,
> 	Niall O'Reilly
> 	University College Dublin IT Services
> 
> _______________________________________________
> mrtg mailing list
> mrtg at lists.oetiker.ch
> https://lists.oetiker.ch/cgi-bin/listinfo/mrtg


More information about the mrtg mailing list