[rrd-users] rrd for application monitoring

[Simon Hobson]

>The pre-processing step would aggregate data for the step size. It
>would compute any desired aggregate data, like averages and
>percentiles. The rrd table would then contain datasources for each of
>the aggregate values.

It doesn't have to compute values for a whole step, as long as it 
performs an update no more often than once a second. You can feed in 
data more often than once/step and rrd will aggregate/normalise it 
over the step.

>I was under the impression
>that you could give it multiple values in the step and it would
>compute some of the aggregate values.

Over one step period, all aggregate functions are the same as the 
value. RRD normalises the data to arrive at a single value for the 
step period - aggregate functions then use only this normalised value.

Eg, if you step is 300s (5 minutes) and the input type is gauge, you 
feed it 5 for one minute, then 0 for four minutes, this normalises to 
1. Max is 1, Min is 1, Average is 1, Last is 1. If you wanted to see 
the 5 and 0 as Max and Min then you would need to use a smaller step 
size and have a 300s consolidation if you still want a 5 minute 
sample to be stored.


-- 
Simon Hobson

Visit http://www.magpiesnestpublishing.co.uk/ for books by acclaimed
author Gladys Hobson. Novels - poetry - short stories - ideal as
Christmas stocking fillers. Some available as e-books.