[c-nsp] netflow sampling

[Jon Lewis]

On Wed, 20 May 2009, Roland Dobbins wrote:

> As mentioned earlier, sampled NetFlow is viewed by many as quite accurate 
> enough for accounting/billing.

If you're only looking at 1/64th of the packets, how do you accurately 
bill for traffic?  Are you assuming that netflow would be collected on 
the customer-facing interfaces, and that the data for each would simply be 
multiplied by 64 (in the case of sampling time-based 64)?

----------------------------------------------------------------------
  Jon Lewis                   |  I route
  Senior Network Engineer     |  therefore you are
  Atlantic Net                |
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________