[c-nsp] Cat 6500 and counters

[Alexandre Snarskii]

On Thu, Apr 28, 2005 at 08:39:40AM +0200, Gert Doering wrote:
> Hi,
> 
> thanks to all that have responded so far.  Seems we'll go the Sup730-3B
> road (and not XL, for now).
> 
> One question remains, though:
> 
> On Wed, Apr 27, 2005 at 02:41:36PM -0700, Kevin Graham wrote:
> [..]
> > > What about netflow?  Does it work on the Sup2/MSFC2 and/or the Sup720?
> > > I saw one post claiming "v5 only", so IPv6 netflow isn't there?
> > 
> > Doubt v9 exports are there yet, but your biggest caveat will be that
> > you'll have to do sampled netflow to avoid killing the Sup cpu. This
> > is my single biggest gripe w/ the platform.
> 
> Can someone else confirm this, that "unsampled netflow" will be a 
> CPU problem?

No, netflow does not have any impact on CPU at all - netflow statistics 
collected at "hardware" level.

> For billing, we won't need netflow on most of the ports, but for some
> parts of the network, we can't do without (yet), and sampled netflow is
> BAD for that (if just because there have been some cases in court recently
> where the court said "this is not science, this is just guesswork, it
> cannot be accepted as a serious base for billing").

Please note that c65 limited on number of active of netflow entries. 
For sup720-3bxl/sup720-3b this number is relatively high one (~230K entries), 
but this limit exists... 

PS: one more limitation - you will not be able to limit router to 
collect netflow data on just some interfaces - it will collect 
data on all interfaces.