[rrd-users] Re: Weird #'s from basic rrdgraph
Alex van den Bogaerdt
alex at slot.hollandcasino.nl
Fri Mar 16 22:22:07 MET 2001
Clarence Hart wrote:
>
> Why do I get different values when I make my start and end times
> over aprox 13days?
> -> If I choose to graph 5 days the max value would be 140 pages.
> -> If I choose to graph 20 days the max value would change to ~70
> pages.
>
> Why does this happen?
If you are working with the AVERAGE values, any peaks in the
data will be flattened out. For instance:
A five-minute sample rate and also 5 minutes per row in the RRA.
t+00:05 value 120 (valid for interval t+00:00 to t+00:05)
t+00:10 value 80 (valid for interval t+00:05 to t+00:10)
t+00:15 value 100 (valid for interval t+00:10 to t+00:15)
t+00:20 value 100 (valid for interval t+00:15 to t+00:20)
If you fill two RRAs using this data, you'll get:
t+00:05 average 120 maximum 120 (valid for t+00:00 to t+00:05)
t+00:10 average 80 maximum 80 (valid for t+00:05 to t+00:10)
t+00:15 average 100 maximum 100 (valid for t+00:10 to t+00:15)
t+00:20 average 100 maximum 100 (valid for t+00:15 to t+00:20)
If this is consolidated into an RRA with 20 minutes per row, and
there is such an RRA for MAXIMUM values and one for AVERAGES, you
get two different numbers:
t+00:20 average 100 maximum 120 (valid for t+00:00 to t+00:20)
Note that the maximum seen is copied to the maximum RRA and the
average is calculated by averaging the averages.
So, you have four RRAs: 5 minutes per row and 20 minutes per row,
averages and maxima.
RRA0: 5-minute averages: 120,80,100,100
RRA1: 5-minute maxima: 120,80,100,100
RRA2: 20-minute averages: 100
RRA3: 20-minute maxima: 120
If you select the wrong RRA, you are selecting the maximum of the
averages and are thus getting 100, not 120.
I hope this makes it clear.
cheers
Alex
--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive http://www.ee.ethz.ch/~slist/rrd-users
WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
More information about the rrd-users
mailing list