On 03/01/2014 11:35, Bharath R wrote:
Hi,
I have a query on cost based formula for finding Extended P-space.
The current formula applies to Node y that is
1). D_opt(Ni, Y) < (D_opt(Ni, S) + D_opt(S, Protected_link.RemoteNode) + D_opt(Protected_link.RemoteNode, Y))

Bharath

I cannot see this text in the draft. Is it from the text or your code?

Also the example below has been corrupted.

Not that asymmetric costs are always problematic in LFA solutions, but
it is not clear how frequently they show in real topologies other than
by accident.

Stewart



I guess the rationale behind the current formula is that the packet on the LDP tunnel should not traverse the ProtectedLink.
But, please consider the following toplogy.
A2àB cost is 5 and A2àS cost is 3; (Both only in the direction as indicated., the other direction cost is unity)
All other costs are unity.
Following the formula, both A1 and A2 can become backup neighbors(with P-Space B and C). However, an LDP tunnel setup via A2 will loopback to S, taking A2àSàA1 path.
  |-------------|
    |     S       |
    |-------------|
      |      |      |
    |      |      |
|      |      |ProtectedLink
    |      |      |
        |      |      |
                                                 | |      |
---------------      |      ----------------
|                       |                          |
|                       |                          |
|                       |                          |
|                       |                          |
|                        |5                        |
|-----------|         |-----------|        |-----------|
| A1 | | A2 | | E |
|-----------|         |-----------|        |-----------|
| |
|                                                |
|                                                    |
|                                                   |
|                                                   |
                               |-----------|                 |-----------|
| B | | D |
|-----------|                                |-----------|
|                                                  |
|                                                  |
|                  |-----------|              |
|--------------|     C      |-----------|
     |-----------|
How about using the following basic formula instead?
2). D_opt(Ni, Y) < (D_opt(Ni, S) + D_opt(S, Y))
Advantages:

  * The formula two ensures that the packet does not loopback to S.
  * Also as per my understanding, if there is Neighbor Ni that
    satisfies 1), but does not satisfy 2), it implies that there is
    definitely one more Neigbhro Nj which satisfies 1) both 1) and 2).

(Because if there is a neighbor satisfying 1) and not satisfy 2), then the packet will loopback via some other neighbor which satisfies 2)).

Given that formula 2) ensures that we don't select any neighbor which simply loopbacks the packet to the source and forwards via other viable Neighbor, it will be more advantageous to employ Formula 2). Please let me know if I am missing something, or if there is a good use case that may prefer backup-neighbor A2 in this example, over backup neighbor A1.(Given that traffic will exit via A1 anyway).
Thanks and Regards,
Bharath R.


--
For corporate legal information go to:

http://www.cisco.com/web/about/doing_business/legal/cri/index.html

_______________________________________________
rtgwg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/rtgwg

Reply via email to