[c-nsp] Sup2T - poor netflow performance

[Roland Dobbins]

"Rolf Hanßen" <nsp at rhanssen.de> wrote:

>This reduces cpu usage of the netflow processes to less than 5%, but it
>looks to me that the exported data is not useful anymore (a stream with
>double of the pps rate than everything else was not recognized by the
>collector software).

Sampling is the normal mode of flow telemetry generation used by large network operators, so it's utility is pretty well-established. Did you configure the collection/analysis software so that it knows the sampling ratio from this exporter, & are you sure that the traffic in question was traversing the interfaces on which NetFlow is enabled in the appropriate direction? 

---------------------------------------
Roland Dobbins <rdobbins at arbor.net>