[rrd-users] Optimize RRD update

[Sam Umbach]

On Feb 3, 2008 12:48 PM, Gwenael Lahay <gwenael.lahay at wanadoo.fr> wrote:
> Hi,
>
> But, if i did not put zero value in the rdd, the rrd tool never calculates
> correctly the average value for 5 minutes and half and hour.
> For example, i have the following value for a 5 minutes interval :
> NaN 3 NaN NaN NaN NaN
> The program store a NaN data for this 5 minutes interval.
>
> Perhaps, i have made a mistake in my creation parameters ?
> It is also possible that the values that i stored are so small that the rrd
> tool calculte an average of zero ?

Gwenael,

One way to alter this behavior is to increase the xfiles factor (xff
when defining your RRAs).  In this case, if the xff for your 5-minute
RRA is set to 0.8 or higher, the "average" across five minutes would
be set to 3 (as up to 80% of the steps are allowed to be NaN, and the
steps without recorded values will be ignored for calculation
purposes).  If you expect the average to be set to 3/5 = 0.6, then you
do have to enter zeros for each step rather than leave them as NaNs.

The NaN values indicate an unknown quantity.  If you are measuring a
recording a zero value, you should enter this into the RRD.  If this
is causing a performance problem, remember that you can collect and
batch several RRD updates to run at once.  Please see the archives of
this mailing list for more information.

Thanks,
Sam Umbach
sumbach at gmail.com