[mrtg] Weired behavior of MRTG

[Alex van den Bogaerdt]

On Mon, Sep 10, 2007 at 05:56:04AM -0700, Peter wrote:

> I think that it's MRTG's problem, because that time every day is the 'peak hour' for my customer's bandwidth... Also, the Threshold check is on his peak traffic utilization... 11.5Gb.. too many coincidence... believe, the easiest thing for me is to say to my boss "no fault found on my mrtg graphs"... but i have to prove it.. :P
> 

Peak hour... as in: a high consumer of bandwidth is present?
that's why I asked:

> > Why would such fluctuations be a problem?  Are there devices
> > with gigabit ethernet (or similar) on the network?


> Well, the problem exists on the aggregated graphs.. if you see the graphs of each 10gig interface separately, it seems that everything is working fine, but on aggregated graph i have this strange behavior! 

Many different consumers will create tiny fluctuations which are
hardly visible.  Also, they tend to cancel each other out.

One consumer with a fluctuating bandwidth demand will create
fluctuations on each of the three channels. When added, those
fluctuations will not cancel each other out but rather will
amplify the result.

When two sinus waves are out of sync, the sum is a flat line.
When they are in sync, the sum is a sinus wave with double the
amplitude.  This is what I'm thinking about.  I could be wrong,
but it is certainly a possibility.

Have you looked at the data in the logfiles and done the math?