TE in Ring Topology

gauravprakash · ‎01-25-2010

HI,

I have 3 Rings in Hiearchy in my MPLS/Ethernet network. Acc, Agg, Core. I want to run Traffic Engineering solely for Fast Convergence & minimise traffic loss due to failures.

RSVP FRR ( link & node protection) is usually explained in context where partial mesh or triangle topology is used.

I read Path Protection is better suited for Ring topology..

Can u suggest a good model to achieve fastes convergence ( sub 50ms or near) & least traffic loss in ring topology network..

Say most of the traffic enters through ACC ring nodes and goes to Core ring and then moves out..

Thx,

Gaurav

Igor Tumkin · ‎01-28-2010

Hi,

If your main goal is fast convergence, go for FRR, since Path Protection is much slower in switchover than FRR (due to propagation time).

FRR may have not very optimal backup path, but it is normally used only for the short time untill the primary tunnel gets recovered.

On the other hand, Path Protection may be better suited for networks, where failures last long and sub-optimal forwading (of FRR) during the failure could be an issue.

Also be aware, that it is 50ms only for locally detected failure. If interface does not go down and the neighbour is just not responding you have to wait BFD timeout or RSVP timeout, which are much bigger.

Thanks,

Igor

gauravprakash · ‎02-03-2010

FRR protects NH ( link ) NNHOP ( node). If I take example of FRR link protection in ring topology with 1Hop primary RSVP TE tunnel and multihop backup tunnel, if the primary TE tunnel fails , the traffic will reroute on backup tunnel in reverse direction ( other side of ring), reach the same next-hop router. IN this case there will be extra traffic reverse and then again towards the destination..

r1----r2----r3---r4----r4-----r5-----r6---r1

Say the ring consists of 6 routers connected via one hop primary tunnels and multi-hop backup tunnel. say traffic with source r1 and destination r5.

now r1 --r2 link fails... the traffic goes into Backup Tunnel (r1--r6--r5--r4---r3--r2) reaches r2 and then again goes from r2 towards r3 --r4--r5, to reach the destination. So, suboptimal back and forward traffic to destination, till the r1----r2 Primary TE tunnel comes up.

Do we agree on this ?

Thx,

GP

Igor Tumkin · ‎02-03-2010

Yes, that's correct.

So FRR is not well suited, if failures may last long.

But if you are really for 50ms switchover there is no other choice.

Regards,

Igor