Re: [hobbit] RRD graphing scales

[Henrik Stoerner <henrik (at) hswn.dk>]

In <759D40D52DC37D4E980D9DC9CEC59880114CDB (at) IRONV.imc2k3.local> "Gregory Wege" <gwege (at) infomanage.net> writes:

>I'm completing a performance analysis of several  servers. In doing this
>I'm seeing inconsistency in the scales in the various charts displayed
>in Hobbit when viewed for different time periods.  For example one of my
>web servers had a performance spike, this spiked was picked up in the 24
>hour graph, but not the 12 hour. Is there anything that can be done to
>remedy these scales?

This is a feature of the way RRD works - data is always averaged over
a period of time. Each of the four datasets contain 576 datapoints, so
the 48-hour graph averages data over (48 hours / 576) = 5 minutes - this
happens to be the same as the polling interval, so the 48-hour graph
will show individual measurements.

However, the 12-day graph (I assume that's the one you're referring to)
still has only 576 datapoints, so each of these is an average over a
30-minute period. This will obviously flatten out brief spikes in your
observations.


Regards,
Henrik