[j-nsp] recommended Netflow sampling rates?

[Justin M. Streiner]

All:

This is aimed more at people who running Juniper routers in fairly high 
traffic environments, but any input is certainly welcome.

I'm looking for recommendations/bast practices regarding what are good 
starting points for setting the Netflow sampling rates to balance 
accuracy of flow stats against not blowing the ASPIC-II out of the 
water...

I've read a bit about this and the consensus seems to be that there is no 
consensus :)  I've seen implementations using between 1 packet out of 
every 100 to 1 out of 1000 with differing run lengths.

We do use Netflow based monitoring, so there is a need to sample often 
enough to keep the data reasonably accurate.

jms