[rrd-users] Rate Of Change question
Marius Rugan
mariusrugan at gmail.com
Wed Mar 19 13:17:35 CET 2008
Hi all,
I was browsing the lists archives from 2003 in search for an implementation
of Rate Of Change calculated on
a GAUGE DS.
what i found is:
https://lists.oetiker.ch/pipermail/rrd-users/2003-November/007513.html
i'm trying to figure out the best logical interpretation of ROC calculated
over datapoints for several timeframes
(day,week, etc). my assumption is that if calculated correctly the number
returned by the ROC would
"trigger" inspecting the graphs visually for something wrong. (for example
a ROC of 8 for a UPS load
or a temperature graph means something ugly is happening down there, and a
normal ROC would be a number that
heads as close to 0 (zero) value )
following the discussion in 2003:
VDEF:first=loadpct,FIRST
CDEF:oldvalue=TIME,POP,loadpct
CDEF:oldtime=loadpct,POP,TIME
CDEF:roc=loadpct,PREV(oldvalue),-,TIME,PREV(oldtime),-,/
printing out the result (on the graph for now)
GPRINT:roc:AVERAGE:Average ROC (Rate Of Change): %2.6lf"
the questions are:
- roc CDEFs results are available as an internal collection of values that
are "willing" to be consolidated
by AVERAGE,MIN,MAX,LAST functions. I didn't found any explanations for the
AVERAGE
consolidating function - is it an arithmetical sum ? e.g. roc1+roc2+roc3 / 3
- is there another way of calculating over the CDEFs results - e.g. my own
consolidation function - sum of square values for example
(i suspect that using the AVERAGE function is misleading the results to 0)
I might have taken the wrong path. if anyone is statistically freak enough
and willing to guide i'm all ears.
thanks in advance for your support
Marius Rugan
--------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20080319/17de53d4/attachment.html
More information about the rrd-users
mailing list