Re: [nsp] flexwan for 6500

From: Andrew Fort (afort@choqolat.org)
Date: Wed Oct 03 2001 - 22:51:25 EDT


On Wed, Oct 03, 2001 at 06:49:35PM -0700, Benjie Ko wrote:

> > we're just using our native ios cat6k as a
> > colo/managed services ethernet switch, so i cant
> > give you any datapoints on the flexwan. cisco and

> actually this will also be the primary purpose of our
> 6509. Have 3548s connect via GE (via ISL trunks) on
> our 6509. Collocs will connect to the 10/100 ports of
> the 3548.

sounds like a typical multilayer triangle design, you'll be able to accelerate your STP/HSRP timers a bit to improve failover performance (and use backbonefast in the core and uplinkfast at the edge, of course).

> Are you doing bandwidth policing on your end? Can you
> give me some hints. Im currently testing per interface
> policer. Im trying to police the vlans to their
> contracted bandwidth (1Mb, 2Mb, etc..).

Yes; on the PFC1 (Supervisor1/1A), you cannot set a PIR (peak rate) in an aggregate or microflow policer; you can only set a CIR. If your QoS requirements only needed a CIR, that's fine. But by your question, you want a PIR set; which means I believe the only other option is "rate-limit" on the VLAN interface (CAR). On the PFC2 (Supervisor2), you can set a PIR, so this should do what you need if you have this platform.

Can some Cat6500-native familiar folk lend some more info on this topic? Does rate-limit scale happily on PFC1/MSFC2 (does it scale better or worse than aggregate policiers)? The mix of router and switch hardware functions is a little confusing :)

> We were also hoping to terminate leased circuits on
> the 6509. That is why i am looking at the flexWAN
> module.

Although FlexWAN is 'pricey', it looks nice to be able to do things this way.. probably chat to your SE about any design caveats/restrictions, tho some others on this list have mentioned FlexWAN in the past, also.

-amf



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:50 EDT