[rrd-users] Dynamic Steps

Simon Hobson linux at thehobsons.co.uk
Mon Apr 22 14:47:47 CEST 2013

Oliver Loch wrote:
>So imho trying to solve this via CDEF wouldn't work as we already lost the exact data that we'd need to calculate it.

I think you may still be missing something important.
RRD is not designed to handle exact data.

It can be fudged to do so under limited and specific situations, but in general it does not store exact values that you put in. First the data is aggregated and normalised to fit into the steps, then it is consolidated to fit your RRA definitions.

But there is also something else you are missing. Your power meter has already averaged the data. Each time it ticks over a unit, then all you know is that since the last tick you have used one unit - but not when you used that unit within that period.
On a domestic supply, I cannot see how you could need to deal with more than one unit/second. To do so would need you to use 3600kWHr/second, which is 3600kW, or 3.6MW.
Realistically you'd struggle to get to something like 20kW, which is only one unit every 180s.

However, RRD will correctly handle multiple updates within one step. So if you did manage to reach that sort of power usage, and you had 10 minute steps, then it would cope with you using 3 or 4 units in one step, and will work out the right average after the data is normalised.

Under low usage periods (when your updates are spaced more widely), you'll just get a long area on the graph showing (for example) 100W consumption for 10 hours - regardless of whether you used 3kW for 20 seconds (boiling a kettle), or 100W continuously (light left on).

More information about the rrd-users mailing list