[rrd-users] Re: Making sure stacked order looks good --- average values

Alex van den Bogaerdt alex at slot.hollandcasino.nl
Mon Oct 25 23:54:58 MEST 1999

dave-mlist at bfnet.com wrote:
> Alex> I just wonder: why do you need to know in advance what you can determine
> Alex> later on from real data?  Once you know that the order is wrong, just
> Alex> move things around.
> Because the order changes.  Some days one counter has high activity,
> other days it's a different counter.
> What I'm considering doing now is looking at the latest archived
> average.

I guess we're both saying the same.

One other thing to think of.  IIRC you reasoned that lower activity
counters should probably be on the bottom because the variance of
higher activity counters might be larger than the lower counters.

This is not always true: I've encountered an example where the variance
in the lower counter made the higher but steady value look *very* ugly.
This is a case where the data flow changes on double the sample interval
and I get an interference pattern on the boundary between medium and high
when graphing <low>,<medium>,<high>.

What about really measuring the variance and decide on that?  You could
determine the variance by looking at max, min and average.  This will
not be too easy since you also need to take in account how many times
you are above or below average.  Perhaps you could get the max and the
average from an RRA with a lower resolution and base your decision on

Other option: don't stack.

In either case, moving the order of your values around might in itself
confuse you and others.

 / alex at slot.hollandcasino.nl                  alex at ergens.op.het.net \
| work                                                         private |
| My employer is capable of speaking therefore I speak only for myself |

* To unsubscribe from the rrd-users mailing list, send a message with the
  subject: unsubscribe to rrd-users-request at list.ee.ethz.ch

More information about the rrd-users mailing list