The MPLS-OPS Archive

[Christopher Poh <tvpoh@essex.ac.uk>]

To Umit,

5/20/02 9:33:18 PM, Umit.Inan@alcatel.com.tr wrote:
>Hello  ,
>
>"Is it possible that all 10K routes (LSPs) make known to interior router
>through lGP + label distribution protocol?"
>
>Even the router,R2 in our example,can handle that amount of routes and
>associated LSPs
>from memory and cpu,switching hardware point of view.
>
>As far as I'm concerned to distribute BGP learned routes into IGP would not
>be an effective design.
>As long as the interior routers support MPLS, they do not need receive any
>of the BGP routes from the BGP speakers.
>A default route information can be injected into R2's routing table with
>the next hop information R3.
>and R2 may bind a single label to all FECS.If I am not wrong this process
>is known as Egress-Targeted Label Assignment.

A more feasible way to cut down all those many routes is by flooding default route through AS-External-LSA into the IGP domain as what you have described. 
Then it would be true that any packet an AS border router receives (from interior router) will be unlabelled one since the packet would have been tagged with 
egress-targeted label. So, the AS border router will need to perform L3 IP lookup for the packet. Seems like we can't eliminate L3 processing along the path 
at all. Is there any better ways to avoid this? Or, avoiding it is not of any much beneficial?
Could it be true that it is allowable for an interior router to selectively request for a particular LSP from AS border router?


To Robert,

>> Is it possible that all 10K routes (LSPs) make known 
>> to interior router through lGP + label distribution protocol?
>
>No. You could theoretically redistribute BGP to IGP but at the same time
>I would recommend to make sure your resume is up to date :)....

I know my question was totally sound crazy, but I just only want to know whether it can be done or not. Thanks for that piece of advice though.

>> >>CP:Thank you for clarifying about BGP peering. What about if an interior
>> >>CP:router (non-BGP peer) would want to establish a connection to exterior
>> >>CP:routes? Would the LSPs for those routes be available to the interior
>> >>CP:router?
>
>What LSP you have in mind ? LDP or TE ? If TE as long as those dst are
>reachable sure no problem. As far as LDP yes the same. After all you can
>label the default route as well :). Redistributing but only specifics
>could be yet another way but I am not sure what application such a LSP
>would be used for.

If it were to distribute only specific routes, is it true that the LSPs are naturally established through control driven triggered by IGP routing protocol. I am just 
curious, please bear with me. :)

Thank you.

-Chris


-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml