[rrd-users] Rrdtool for counting irregular events

[Simon Hobson]

Rafal Gwizdala wrote:
>That's a little difficult. My application allows the users to 
>specify date range themselves by dragging a selection over the graph 
>(just like in Cacti)

Ah yes, that's a bit difficult then.

>So I don't have a control on what the date range will be.
>Anyway, would you mind telling us about your best practices for 
>graph range/step calculation?

Well I only have four consolidations in my databases - designed to 
for for day, week, month, year graphs. Thus it's simple to 
pre-caculate the optimum resolution and timespan values for each 
graph. I just use a case statement to pick one of four sets of values 
and plug them into the graph generation script.
I carefully choose the timespan of the graph so that each pixel on 
the graph exactly matches on consolidated data point.

If you choose any other combination then you won't get more 
information in the graph - unless you are storing high res data for a 
longer period. So for the consolidations I've chosen, if the user 
were to select (say) a 15 day period to graph, then the best they'd 
get would be the same data as used for a 31 day graph - so about 2 
pixels per data point.

On the other hand, if you store more data (eg 1/2 hour consolidations 
for a month instead of a week) then by selecting a 15 day window the 
user would then get more information on the graph. There is a lot of 
flexibility to trade off storage space vs reporting ability there.

-- 
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.