[j-nsp] IP/MPLS fast convergence

[Mark Tinka]

On Thursday, December 22, 2011 05:20:30 AM Phil Bedard 
wrote:

> As for the design it really depends on the SLA.  If you
> do not need 50ms Traffic restoration LDP should work
> fine.  On a metro network IGP convergence is pretty fast
> these days probably less than 500ms.

Indeed.

We run RSVP-TE only between the Aggregation routers (Juniper 
MX480's, M320's, T320's), with the core just processing 
those messages accordingly. As this is where our IPTv 
services terminate, it makes sense for us.

We don't run RSVP in the Access (Cisco ME3600X's), just LDP. 
This runs between the Access switches in a ring, and 
terminates on the Aggregation routers at either end of the 
ring. We then tunnel LDP inside RSVP across to other rings, 
although the Aggregation routers also run LDP on all 
interfaces.

We've simulated failures in the live networks across 
different rings and also within the same ring. And yes, IGP 
convergence can fall between 300ms - 750ms, give or take. We 
run BFD throughout the entire network, too. IGP is IS-IS.

It is likely we could support RSVP-TE in the Access, but 
this would be a commercial decision, i.e., a customer 
requires 50ms failover across their protected l2vpn/l3vpn 
service - but this would be quite pricey to discourage the 
practice, as running RSVP is quite hectic.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20111222/52f82108/attachment.sig>