[c-nsp] High Availability PoP Design

[Mark Tohill]

Thanks to all for your suggestions.

We have moved down the road of meshing between switches and routers, and
running BVI's across GigE ports on each of the routers, - not my choice.

Etherchannel between S1 and S2 now blocking at S1 and resiliency works
on loosing switch. :) Just now trying to remind myself why STP blocking
where it is, so when we double up on the routers, we have a
predetermined, predictable path, hopefully.

Thanks again to all,
Mark.

-----Original Message-----
From: David Coulson [mailto:david at davidcoulson.net] 
Sent: 02 November 2005 16:48
To: Michael Markstaller
Cc: Mark Tohill; cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] High Availability PoP Design

This isn't perfect, and has some caveats (specificly SSO and performance
related), but we do this:

7507 with a pair of PA-FE-TXs. Each PA-FE-TX on a router is connected to
a different switch, which are then interconnected.

Each VLAN is a sub-if on the FE and we do:

bridge 100 proto ieee
bridge 100 route ip

int f0/0/0.100
encap dot 100
bridge-group 100

int f6/0/0.100
encap dot 100
bridge-group 100

int bvi100
ip addr 192.168.1.2
stand 100 prio 110
stand 100 pre
stand 100 ip 192.168.1.1

Same config on both boxes, except the IP of the BVI is different. Seems
to work pretty well so far. Best L2 meshing technique I can come up with
without breaking a whole lot of other things - Running an IGP on servers
isn't really an option.

David

-- 
David J. Coulson
email: david at davidcoulson.net
web: http://www.davidcoulson.net/