[rrd-users] More observations and questions on COUNTER

Simon Hobson linux at thehobsons.co.uk
Sun Oct 24 18:58:20 CEST 2010

Philip Peake wrote:

>So you are saying the rate is always per second?
>I thought it was rate per <sample period> ??

It is always <somethings> per second.

>So a value stored at the exact end of a sample period would be equal to
>the value presented.

IF (and only if) you feed in updates exactly on every step boundary 
(and remembering that the refernce is always midnight 1st Jan 1970 
(unix epoch), then yes, the value stored for that step will be what 
you feed in - subject to any conversion required (in this case 
subtracting the previous counter value to get the difference). Under 
any other circumstance you will get normalisation.

This is a commondly asked question, and the simple answer is that RRD 
Tool is not designed to store original values - it is designed to 
store rates. Any fudges to make it store accurate values are just 
that. Also, rates are always stored as a floating point number.

See Alex's tutorial on normalisation and consolidation at 

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.

More information about the rrd-users mailing list