[rrd-users] icmp measurements with rrd?

itris at technologist.com itris at technologist.com
Wed Nov 1 02:01:27 MET 2000

```greetings,

I've been using RRD for a while now and am quite happy with it :)
Recently I received a request from my employer to graph the latency and
packetloss in our network. My first choice would be MRTG, but due to
it's limitation to be able to graph/measure only 2 things at a time
(leaving out max/min and any other things I'd like to graph/measure) and
some problems with scaling it we probably won't use it.

So then the question becomes, RRD or build your own.
I'd like to try RRD.

But I have a few questions first that I'd hope to get an answer to:

Let's say I have an RRD with a 5 second measurement interval. The
measurement will take some time, but it'll be less than a second. So
we'd be putting data in the RRD at time+0+measurement_time,
time+5+measurement_time, etc..
Would then, a measurement_time of 0.3 (and hence insertion time of for
instance time+5.3) be subjected to resampling, resulting in a
measurement of (if the measurement was 0.1) (0.1/5.3)*5 = 0.94 ?
Or, would the time+5.3 insertion time be too late for the step, and be
taken as time+6, resulting in (0.1/6)*5=0.83 as measured value?

I suspect both cases will be answered with "no".
The time needed to measure the latency is dependent on the latency
itself.. So if for some weird reason the latency would rise above 1
second, the measurement would take more than 1 second.
Therefore, the insertion time would be time+6.3 (for instance) with a
measured value of 1.2
The RRD will therefore resample and show a value of (1.2/6.3)*5 = 0.95
If this is correct, is there any way around this ?
(other than measuring every second, eliminating any resampling, but no
longer allowing latency to be measured over 1 second and creating so
much network traffic that it will influence the measurement)
Otherwise I think RRD will be useless for measuring latency..

Itris

--
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