[rrd-developers] How to create a new Data Type and Consolidation Function?

[Long V]

----- Original Message ----

> From: Dan Cech <dcech at phpwerx.net>
> To: rrd-developers at lists.oetiker.ch
> Sent: Tue, April 12, 2011 2:02:25 PM
> Subject: Re: [rrd-developers] How to create a new Data Type and Consolidation 
>Function?
> 
> Back up a second, for a regular COUNTER the rate is just the raw 
> difference divided by the time delta, so you can construct your 
> cumulative difference data from the AVERAGE easily with a  CDEF.
> 
> Assuming a 60 second step:
> 
> time            ___0 ___1 ___2 ___3 ___4 ___5 ___6 ___7 ___8
> counter_value  ___0  _300 _450 _450 _900 1200 1800 2100 2550
> rate1                ___5 _2.5 ___0 _7.5 ___5 __10 ___5 _7.5
> rate1 *  60          _300 _150 ___0 _450 _300 _600 _300  _450
> rate2                   3.75      3.75     _7.5        6.25
> rate2 * 120            _450       _450     _900       _750
> rate3                        3.75                 6.875
> rate3 * 240                  _900                 _1650
>

Hi Dan, you are correct. 

But using a CDEF to alter the data implies you need to know which RRA will the 
time range trigger (when you want to zoom-in the data).

Example: assume resolution 1 we keep 4 PDP. Resolution 2 we keep 4 CDP. Then if 
our zoom time range is time 5 to 8 then it'll trigger res_1 so we have to 
multiply by 60. If our zoom time range is time 1 to 8 then it'll trigger res_2 
and we have to know it's res_2 so we have to multiply by 120 instead of 60.

With Cacti, we can only associate one single CDEF per DS no matter how many RRA 
this DS has. The CDEF is static. So already with multiple RRA, we need to at 
least have one corresponding CDEF per RRA to re-multiply by the proper step size 
of the RRA.

When zooming, Cacti just generate the time range and feed it to rrd_graph. Cacti 
can also pass a CDEF to rrd_graph. But there's only 1 CDEF per DS. Even if we 
had one CDEF per RRA (so many CDEF per DS), Cacti still has no way of knowing 
which CDEF to give because it doens't know which RRA will be triggered by the 
time range.

Long

> 
> What am I missing  here?
> 
> Dan