The MPLS WG Archive

["David Allan" <dallan@nortelnetworks.com>]

Hongwei:

 

You use the tools according to the requirements. Ingress router re-route on failure will of course make the most optimal use of resources as the only network resource committment is to the currently working "optimal" path. However on failure you get a relatively long outage which is the sum of the durations of the detection/propagation times for the initial failure, and the set up time for the new optimal path with the failed resource removed from the topology database. Any attempt to improve on this to meet other performance criteria (e.g. to minimize packet loss) involves juggling how you commit resources in the network and steady state operation may look less optimal if you simply stick with any one approach.

 

As to your questions, won't claim expertise... but an explicit route (and I assume you mean with constraints) involves pruning the topology database according to the desired constraints and running the SPF algorithm on the pruned set. How much computational load this spreads over the network (e.g. done once at the LER or iteratively modified as the setup progresses across the network) depends on whether you use a strict or loose route, have routing hierarchy, rate of churn, etc. crankback being one answer to fix imperfections in LER route determination.

 

later

Dave

-----Original Message-----
From: Hongwei [mailto:hongwei.zhou@elec.qmul.ac.uk]
Sent: Tuesday, October 09, 2001 2:28 PM
To: Allan, David [CAR:NS00:EXCH]
Cc: mpls
Subject: RE: LSP Failure/Recovery

 

Dave,

 

If it is intended to implement traffic engineering with mpls, i think it will be rational to let the ingress router deal with rerouting task because it is the ingress router that computed the explicit route. if we allow LSR16 to do local rerouting, we may not achieve overall optimised network performance, that is, the network load may not be balanced the best.

 

If we have no concern about traffic engineering, i think all the three methods you listed are usable, depending on the network administrator or device vendors.

 

To be honest, I have also some other confusions about explicit routing:

1. how does the ingress routers compute the optimised explicit routes through the network?

2. To compute an optimised route, a router need complete information of the network. but every router including LERs and LSRs, has the complete information about the whole network or the autonomous system, or just the LERS have this whole information?

3. how do the routers get the information about the whole network? it should be a headache because the load of the network, maybe also the topology of the network, is in highly dynamic change.

 

Thanks for every answer in advance!!!

 

-Hongwei

 

-----Original Message-----
From: owner-mpls@UU.NET [mailto:owner-mpls@UU.NET]On Behalf Of David Allan
Sent: 09 October 2001 18:39
To: Sachin Kalra; Naidu, Venkata; 'manis@futsoft.com'; mpls@uu.net
Subject: RE: LSP Failure/Recovery

Sachin/Venkata:

 

You're mixing protection switching, and re-route scenarios. There are specific behaviors/state requirements depending on the recovery strategy,

    - LSR 16 may tell the ingress (revertive/non-revertive global repair),

    - LSR 16 may tell no-one (local repair), 

    - LSR 16 may tell EVERYONE (simple re-route) when the link goes down/comes back up.

 

later

Dave