[mrtg] Re: MRTG Graphs - True CIR representation
pwilliamson at mandtbank.com
Thu Feb 2 19:46:55 MET 2006
>>> Steve <crunkmeister at gmail.com> 02/02/06 1:35 PM >>>
> Hello All,
> First off I want to say thanks in advance for any help I recieve!
> Second, I want to say I am no perl expert, I have fumbled
> through getting MRTG to work for me but I have little experience
> with commands and such. My question is: I am monitoring T1,
> Frame-Relay, CPU and Ethernet traffic. The graphs only grow
> to the maximum bandwidth being used. The grow to the size of the
> spikes, they are not a true representation of the actual "pipe" ie.
> A T1 would be 1.544 mbps or a 512 frame relay. I'd like the
> graphs to "always" stay at here true commited rate and have
> the actual data benn represented within the graph. I hope this
> makes sense. Thanks for the help!
You have a few problems wotrking against you here...
1. CIR is measured per second, so that you're guaranteed
at least that, but you can always go over. We have that happen
to us all the time, our vendor let's us leak over the CIR whenever
we need to it seems like.
2. MRTG is polling in 5 minute intervals. It is not a detailed
diagnostic tool like a sniffer. MRTG will definitely miss the peaks
and valleys of a very "bursty" network. We have one of these
as well, and have decided that the "burstieness" is something
we live with, but will catch any traffic trends over the longer
So, you do get the actual data within that polling period. You may
have exceeded that CIR 8 times during the polling period, but
it may never show on the graphs. Especially if you're using
MRTG and polling at 5 minute intervals. You could probably do
a bit better with RRDTool and polling more frequently, but then
you increase the error with which RRDTool will interpolate
the values collected.
Or, just add Unscaled[ezwf]:ymwd to your configs.
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
FAQ http://faq.mrtg.org Homepage http://www.mrtg.org
More information about the mrtg