[rrd-developers] DS idea

[Tobias Oetiker]

Hi Peter,

Today Peter Grotrian wrote:

> Hi
>
> I have just started using RRD and mostly it works just great.
>
> I have an existing framework which samples data from a system
> every minute. To avoid load peaks, the sampling of the different
> parts is spread around the seconds of a minute. For most measured
> variables RRD does just the right thing by interpolating to the
> step grid.
>
> However, part of the collected samples are maximum wait times of
> queueing systems. These numbers are not guesswork but calculated
> exactly while servicing. I am very much interested in seeing the
> maxima of these maximum wait times later when I query a MAX-RRA.
> Currently, RRD interpolates to the grid first and than computes
> the maximum.
>
> I know I could set the step size to 1 second and no interpolation
> would happen. This looks ugly to me, as I sample by the minute
> and would have to set very high XFFs. I could also fake update
> times by the exact minute but that would adversely affect the
> non-maximum samples. I could use separate RRDs for queueing
> system maxima and the rest but I'd rather have one RRD per
> observed system module.
>
> Why do I post this on the developer list?
>
> I would try to implement a new DS, let's call it MAXGAUGE. This
> DS would only be allowed for MAX-RRAs. Interpolation would not
> happen but rather the update value would placed at the latest
> step at or before the update time.
>
> What do you think of it?

sounds like a sensible analysis to me :-) except that I would call
the datasource type

  GAUGELATEST

cheers
tobi





> Regards, Peter
>
> _______________________________________________
> rrd-developers mailing list
> rrd-developers at lists.oetiker.ch
> https://lists.oetiker.ch/cgi-bin/listinfo/rrd-developers
>
>

-- 
Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten, Switzerland
http://it.oetiker.ch tobi at oetiker.ch ++41 62 775 9902 / sb: -9900