[rrd-users] More observations and questions on COUNTER

[Philip Peake]

On 10/24/2010 11:59 AM, Simon Hobson wrote:
> Philip Peake wrote:
>
>> So if you look at the data I present to RRD, you will see that it is on
>> 30 second boundaries, so I would expect the stored value to be the same
>> as the presented value -- no ?
> NO !
> You are using a counter data type, therefore your data WILL be 
> processed before storage. The calculation is (current - 
> previous)/time = rate
>
> Eg, store 1000 at time 0, store 4000 at time 300, stored value is 
> (4000-1000)/300 = 10
>

Ok - you are saying that its always stored as a rate per second.
For some reason, I thought it was rate per sample-period.

That agree with what I am seeing.

Philip