Hi all,

I have been using MRTG and RRDtool for many years now, and now I am considering using RRD for data from electricity meters. Does anybody use RRDtool for this purpose? What is your RRD file configuration? Some challenges I see with electricity meters as opposed to ifInOctets-like network traffic data:

• So far I think I will use step of 900 seconds, but will read values more often (say, every minute or every 5 minutes), based on whether somebody is looking at them right now, and I would like to use the more precise data for e.g. power peaks calculation: If I read $e1 at $t1 and $e2 at $t2 (where $t2-$t1 is much smaller than $step), I would like to use ($e2-$e1)/($t2-$t1) as a possible power peak, not the average power over the whole $step.

• I use mostly MODBUS/RTU-based meters, and reading all the meters on one RS485 line takes about 15-20 seconds. So for interactive data, I want to store the exact timestamp of each reading (per-meter). This probably means a dedicated RRD file for each meter.

• The energy meter is essentially an ever-increasing COUNTER (or DCOUNTER), but I need absolute values as well - for example, I would like to be able to calculate the exact energy consumption since YYYY-MM-01 00:00 till YYYY-$(MM+1)-01 00:00, no matter how many data points in between are missing. The LAST RRA would be handy for this, but then I would need absolute values, not rate-transformed COUNTER values. And preferably an exact timestamp of the last measurement in the aggregation interval, not the end of that interval.

• The billing by the electricity distribution company is done monthly here (midninght the 1st to-midnight the 1st), and I would like to be able to match it as close as possible even with my subordinate energy meters. The problem is the daylight-saving time - usually April is 30×86_400-3600 seconds long here, and October is 31×86_400+3600 seconds long. But RRA with 1d aggregation works in multiples of 86_400 seconds even across the daylight saving time change, doesn't it?

• The electricity meters usually provide the energy consumption in (kilo)Watt-hours, but the rate-transformed unit should be (kilo)Watts, i.e. ($e2-$e1)×3600/($t2-$t1) where $t1 and $t2 are in seconds. I guess some RPN magic can do this :-)

How do you solve the above problems in your systems? I think some of it can be partly solved by storing the electricity meter reading twice, to two DSs - one as GAUGE, and the other one as COUNTER.

Any other suggestions? Thanks!

-Yenya