[rrd-users] Usual RRD question on AVERGAE, MAX and consolidation :)

[Johan Elmerfjord]

Hi James,

Yes, this is exactly how I understand it.

So this is why average is normally used, then it doesn't matter if consolidations happens or not, as it will always average the values that are grouped together.
The peaks and dips will off course be "smoothed", but the real amount that has been transferred should stay about the same.
This is not the case when using MAX.

Some Network-traffic-providers use 95-percentile for the billing.
There are functions to get that built into RRD as well, and it's returned as a number for the interval you are looking at.
I think that the same consolidation happens here as well, so do use Average.

It's also possible to export the data (using fetch) and add up all the values in the interval.
That's a way to get around the consolidation if it's really a problem - without creating some very wide graphs).
Rrdtool xport can also create some useful output where it processes the data as when doing graphs.
Do remember to specify the maxrows or consolidations will occur here as well.,

Good luck,
/Johan



On Thu, 2013-12-19 at 16:35 +0000, James Bensley wrote:


Hi Johan,

Many thanks for the reply. Does RRD draw graphs as one CDP per pixel?
As a simple example; to produce a month graph taking samples every 300
seconds (5 minutes) I would need a graph 8640 pixels wide; 30 (days in
a month, or 31) * 24 (hours in a day) * 12 (5 minute samples in one
hour) = 8640, is that logic correct?

If I am producing graphs say 864 pixels wide then we are consolidating
10 CDPs into each pixel, in this instance when using
DEF:ds0=my.rrd:ds0:MAX we choose the highest of the 10 samples to be
used in that one pixel?

I am graphing with RRD for billing customers based on traffic usage so
I want this value to be as high as possible (without being higher than
what they actually used), otherwise I am under-billing.

Kind regards,
James.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20131219/964cfbfc/attachment-0001.htm