The MPLS WG Archive

["Venkat Dabbara" <venkat.dabb@wipro.com>]

hai abhay ,

            Thanks for the reply ...Well , i too don't see
         ne good reason for having a new next hop availablity 
        in CRLDP . Just wanted to knoe that if there is any 
        standard way in which a LSR should behave when it 
        comes accross a situation ...

         I feel route pinning every CRLSP by default would 
        solve things wherein u never respond to a new next hop
         availablity and continue to use the same old pinned
        path ....Is that ok ??

thanks and regards
venkat
                    

        

On Tue, 20 Nov 2001, abhay wrote:

> hello venkat,
>                  Yes,why not,there is no hard based rule for not storing ER-TLV's
>                    Ref T.E MIB
>                   One reason for not doing so would be that usually Loose hops
>                   are not directly connected.
>                   In other words the administrator does not know the stable topology
>                   and gives it a shot by specifying the next hop to be loose,which means
>                   the loose hop can be reached through any router.(Here if route pinning is
>                   applied,then a better next hop is no use).
>                   
>                   
>                   But,to do that,your MPLS management should very very efficient
>                   In such cases,when you might have a lot of communication between
>                   your signaling and forwarding management.
>                   
>                   Also,there arises a question,of controlling the child like Ingress of the Initiated LSP
>                   to sustain the paths,since this is a downstream on demand mode
>                   
>                   Also,whats the used of the better next hop ?.I dont see any good use of it,when one
>                   has a CR-LSP.
>                   I guess it more of engineering and adding an extra rear view mirror to CR-LDP
>                   Hope this helps,
>                   Regards,
>                   Abhay
> 
> 
> ----- Original Message ----- 
> From: Venkat Dabbara +ADw-venkat.dabb+AEA-wipro.com+AD4-
> To: +ADw-mpls+AEA-UU.NET+AD4-
> Sent: Monday, November 19, 2001 5:01 PM
> Subject: next hop change in CRLDP 
> 
> 
> hai ,
> 
>          This is regarding next hop change in CRLDP ( Better next hop avalability ) .
>          I could find some discussion happening before on this but it wasn't satisfactory 
> 
>          Some one had asked 
> 
> +AD4-Suppose a CRLSP is setup with no-pin option and  LSR-A is 
> +AD4- part of LSP as 
> +AD4- loose hop member. 
> +AD4- When LSR-A finds better nexthop, it is supposed to taer down 
> +AD4- LSP along old 
> +AD4- path  and send label request to new nexthop. For this, LSR-A 
> +AD4- has to store 
> +AD4- all the TLVs it received in CRLSP initiation request so that 
> +AD4- it can initiate 
> +AD4- request along new nexthop. Does not it imply that CR-LDP is 
> +AD4- not scalable? 
> 
> And some one had answered it saying 
> 
> +AD4AIg- This is not a scalability issue but a simple engineering issue, 
> +AD4- no-pinning with a CR-LDP loose hop is an unlikely scenario. 
> +AD4- No pinning is most appropriate with LDP. +ACI-
> 
> Ruling out such a scenario doesn't seem  to solve the problem ..What if the scenario comes in ??
> Surely we can react as it is being done in LDP where u just tear down the path .
> Some thing of +ACI- make and break+ACI-  policy should be done which would be of some meaning rather than going for
>  +ACI-break and make +ACI-
> as done in LDP where u land up tearing down the CRLSP for every better next hop avalibility.
> 
> As pointed out by some one , will it be scalable to store the list of remaining hops present in ERTlv
> at every hop ??? If yes ,this would solve the purpose to some extent ...
> 
> Please correct me if i am wrong.
> 
> regards
> venkat
> 
> 
>         
> 
>       
> 
> 
> 
> 
> 
>