[nsp] rapid pvst and Intel Pro/1000 server adapter teams

[John Ferriby]

In our current configuration we are using Rapid PVST
between two 3750Gs, and each switch has a number of
servers.   We're looking to configure several servers
using Intel Pro/1000 adapters using their adapter teaming
in what they call "Switch Failover Mode".  In this
scenario each adapter of the team targets an l2 switch
(vlan in our case) and can detect and failover in the
event of a switch failure.

Intel assumes in their doco. that regular STP is in use
and recommends a 2x forwarding delay factor for the
forwarding delay on their adapter teams.   This seems
silly for a Rapid PVST environment, where we can see
nearly sub-second cutovers.  (We're also running
UDLD aggressive on the trunks.)

I'm thinking something on the order of 3 or 4, conservatively,
for this parameter.   Does anyone have any real-world
experience with this type of scenario?  Any opinions?

-John