[rrd-users] Re: update < 1s ?

[Iain Buchanan]

I've been using the great graphing properties of rrdtool successfully for a
while now with very frequent data.  The problem is sometimes the data is too
frequent!  ie, more than once a second.  (I have accuracy to about 100ms)
I've got around this by averaging all samples in the same second before I
pass it to rrdtool.  This is ok for the most part, as graphs don't need to
be that accurate.  Except for the GPRINTs which get the max and min over the
time period.  Sometimes I loose an absolute maximum, because it's been
averaged on the same second with some other data.

I am just keeping one DS with as great an accuracy as possible for the
graph, and storing three RRA's: AVERAGE, MAX, and MIN.  I could keep three
DS's, and instead of just averaging 'greater than once per second' data, I
could keep an average, max and min, but would this triple the size of the
rrd?  Am I right in saying that if I define three RRA's for three DS's, (min
max & ave) then rrd will keep three sets of values for every DS, in effect
keeping min of max, ave of max, max of max, min of ave, ave of ave... etc?
Alternatively I could keep three rrd's, each with one DS and one RRA with a
CF matching the DS.

Does anyone have a better suggestion?!

Thanks in advance
Iain



--
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