[mpls] Re: New Version Notification for draft-li-mpls-mna-entropy-03.txt

[Tony Li <[e-mail]>]

Hi Loa,

Your diagram arrived here badly mangled.  I’ve changed to a fix width font and taken the 
liberty of trying to repair it.  Apologies if I’ve only broken it further.

>                     +-----------------------------------------+
>                     |                                         |
>                   +---+                                       |
>            +------| D |                                       |
>            |      +---+                                       |
>            |        |                                         |
>          +---+      +---------------------------------------+ |
>    +-----| B |                                              | |
>    |     +---+      +-------------------------------------+ | |
>    |       |        |                                     | | |
>    |       |      +---+                                   | | |
>    |       +------| E |                                   | | |
>    |              +---+                                   | | |
>    |                |                                     | | |
>    |                +-----------------------------------+ | | |
>    |                                                    | | | |
>  +---+                                                 +-------+
>  | A |                                                 |   Z   |
>  +---+                                                 +-------+
>    |                                                    | | | |
>    |                +-----------------------------------+ | | |
>    |                |                                     | | |
>    |              +---+                                   | | |
>    |       +------| F |                                   | | |
>    |       |      +---+                                   | | |
>    |       |        |                                     | | |
>    |     +---+      +-------------------------------------+ | |
>    +-----|C  |                                              | |
>          +---+      +---------------------------------------+ |
>            |        |                                         |
>            |      +---+                                       |
>            +------| G |                                       |
>                   +---+                                       |
>                     |                                         |
>                     +-----------------------------------------+
> 
> 
> Not a network model, but a description of load-sharing option. You say that
> we need the number of paths that a packet may take through the network. In the figure above a packet coming into node from the can be sent either "north" or "south". If it sent "north" it will enter node B and again go north or south. Once a packet reach node D, E, F, or G, it traverses a part of the network that does not do load-sharing. (Yes, I'm lazy!) To me the load-sharing decision seem to a local matter, what is the impact of the fact that there are 8 routes through the network? What seem to have a bigger impact is the number of out going load-sharing interface, but again that is local.

I’m assuming here that the traffic flows from A to Z.

Yes, load sharing is a completely local decision.  The goal is for each system that has multiple paths to select an appropriate next hop,
consistent with the overall path and avoiding packet reordering within a flow. Since intermediate nodes do not have a clear understanding
of the flows within an LSP, the Entropy Label/Value gives a hint to downstream nodes about how to forward particular flows. If done
correctly, packets with the same Entropy Label/Value will follow the same path and thus avoid reordering.


Tony