[rrd-users] Temperature Monitoring

[Phil White]

Hello.

After a couple of years using RRD to store my temperature sensor logs, I
lay in bed the other day and cursed, after deciding that I was doing it all
wrong...

To cut a very long story short, I am now of the opinion that I should not
set up my RRD store as
  DS:$sensor:GAUGE:300:-50:100
 simply because this is not the information that I am *really* interested
in - it is just the way that most people do it, because that is the way we
are used to thinking.

I'll try to explain a bit better. I have several sensors - some inside the
house, some outside. When I started, the aim was to monitor the
temperature, and use this to automate heating. But, rather than the
*actual* temperature, I am actually more interested in the *relative*
temperature difference between the average outside temperature, and inside
- together with the rate of change in temperature (which are probably
related in some way).

Therefore, I'm looking at redoing the whole lot from scratch. Questions so
far:

DS:delta:DERIVE:300:U:U
appears to support only integer values. Is this a bug, or a design feature?

Since I have a number of sensors, and wish to evaluate the pseudo-equations
  delta=mean(inside) - mean(outside)
and
  the differential rate of change of temperature
does anyone have any opinions / advice on the best way to store the data,
and how to calculate this? I have had a few attempts, but am getting
nowhere fast on these.

I have been using one RRD file per sensor - but I do not necessarily have
to do it this way (though it is easier at the moment). I am looking at
efficiency of CPU and disk I/O, as well as ease of gaining access to the
computed values I need. I could continue recording temp, and then
evaluating the required figures each time - but it seems sort of wrong...

Best regards,

Phil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20120215/6a60417d/attachment-0001.htm