Hi, We have a medium sized network (about 25 sites). In our case we decided to do the following:
On 2 March 2010 07:14, My Name <[email protected]> wrote: > What are everyone's thoughts on redundant route processors and linecard > diversity for P and PE connections. > Specifically, assume redundant P and PE devices, > > 1) do you deploy redundant route processors per node? Yes, in every P or PE we have redundant RP. > 2) do you use one linecard for all P connections (assume medium sized WAN) > or do you install multiple linecards and diversify P connections? How > many? We have up to 4 different paths between the Ps (N, S, E, W). Each of the connections is on a separate line card. If we have multiple links in particular direction (let's say 2xN) then both interfaces are on different line cards. > 3) Same as 2) but for PE connectivity? Yes, but with a variation - In biger PoPs we have two Ps - then the PEs are connected in a 'box' shape (P1 - PE1 - PE2 - P2). If there is only one P we have at least 2 line cards to connect PEs. We tend to mix P and PE connections on the same line cards. > 4) Does BFD failover for all potential failure scenarios eliminating need > for diverse linecards > for port assignment? BFD can only detect failures. If the linecard is dead then if there is no physical redundancy in place then obviously there isn't much you can do anyway. > > I assume the Core should be highly available and you should diversify as > much as possible, > within reason of course. If I had 4 connections for P to P connectivity I > would use at least two > linecards regardless if I had a redundant router or not.Thoughts? It depends on the location of the Ps and the paths between them. If all the connections use the same path - then I would say 2 is enough. In some cases when there are diverse paths and you want to maintain capacity in case of a failure - I'd say more might be needed. In general. If the traffic is important to you and you have to maintain not only connectivity but also quality - more is better (usually for a price). kind regards Pshem _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
