[j-nsp] MX sampling: option run-length

[Justin M. Streiner]

On Sun, 24 Feb 2013, Piotr wrote:

> I'd like ask about this options, I made some test sampling with nfdump+nfsen 
> and i don't see the difference wheter this option is on or off. What is 
> exactly advantages and disadvantages of this option in practise ?

If you do, for example, 1 out-of 100 sampling with a run-length of 4, you 
will sample one packet out of 100, plus the next 4 packets, so you're 
functionally sampling 5 packets out of 100.

The reason your nfsen graphs don't look any different is because 
nfsen/ndfump don't know what to do with those extra 4 packets.  If you 
want 5 out-of 100 sampling to show up in your graphs, you'd need to 
changing your sampling rate to 1 out-of 20, and not use the 'run-length' 
feature, and then change the expected sampling rate in nfsen accordingly.

jms