[c-nsp] Cat 6500 and counters

Dale W. Carder dwcarder at doit.wisc.edu
Thu Apr 28 11:27:54 EDT 2005


On Apr 28, 2005, at 8:38 AM, sthaug at nethelp.no wrote:
>>>> Can someone else confirm this, that "unsampled netflow" will be a
>>>> CPU problem?

If you're worried, just monitor the cpu loads (there are
two cpu's) and netflow entry count via snmp and set up some
thresholds to page you.

>>> No, netflow does not have any impact on CPU at all

This is incorrect.

You will see an increase in switch processor (not route processor)
cpu utilization as a function of netflow entries.

One caveat I ran into is that on the supervisor 720, netflow
entries are mistakenly created for flows that are dropped by
uRPF.

At some point, as much as everyone would love to use netflow for
billing and everything else, on a busy aggregate link it is like
trying to drink out of a fire hydrant.  Sampling is there for a
reason, but may or may not be for you.

Dale



More information about the cisco-nsp mailing list