How long does a redundant pair of 3550s take to failover under HSRP?

The failover isn't a function of the platform, it's a function of the protocol.

By default, HSRP hellos are sent every 3 seconds. The default dead timer is 10 seconds which means a failover (worst case) of  seconds.

The timers (hello and dead) can be changed. It is recommended that the dead timers be set to three times the hello interval.

Okay so if the hello is set to one second and the timer to three the total failure should be around four seconds. However, is it really?

Has anyone tested it to find out the actual time? I keep wondering this because the 6500 can actually do it in the sub 2 second range, where 4000s can take forever.

I very much want to keep the connections alive and whatever setup I end up with will probably be replicated a number of times (I 'm a big believer in heavy standardization).

For the 6500 failover time I was talking about one 6500 with two S2 supervisors

Are you running Dual Router Mode or Single Router Mode?

If DRM, I've seen people crank the timers way down (300ms hello and 1 second dead).

If SRM, HSRP is irrelevant since SSO handles the redundacy.

True, but again I was trying to determine the times for 3XXX managed switches.

I did quite a bit more digging and it is possible to get it down into the under 5 second range with HSRP and EIGRP with a only a few gotchas.

I guess we'll leave it there for now. Thanks for the help.