[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE:Comments on ARO: ARO and PCE order of computation.

Hi,

At 10:22 AM 5/17/2004 +0200, ricciato wrote:
Dear JP, Adrian, CCAMPers,

I have been thinking to the PCE proposal.

Aside from the details of the signaling procedure, it seems to me that
a very basic difference between the PCE
proposal and the ARO proposal is in the order of computation.

In the ARO proposal, the path sections in AS(k) are computed *before* the
path sections in AS(k+1). Instead, in the PCE approach the path
sections in AS(k) are computed *after* the path sections in AS(k+1) (at
least, this is what I understood from sec. 6.6 of your draft).

I've called this approach "egress-influenced path comp. " in my reply
to Adrian, see http://ops.ietf.org/lists/ccamp/ccamp.2004/msg00474.html.
(I relate this to the comment previously issued by Adrian, about the
possibility to let the downstream ASBR to take part to the path computation
process.)

In other words, we can say that the path computation with ARO proceeds
forward, while with PCE it proceeds backward.

Do you confirm this?

Not at all ... the example provided in the draft shows a backward procedure but one could also use a forward path computation procedure. Moreover, there are many implementations details that are not discussed in the draft.
The fundamental difference is that with the PCE-based approach one can built a shortest end to end path (along with other path attributes such as path diversity)

JP.

If the comment above is correct, could you (or Adrian) clarify the
technical reasons why a "backward computation" approach is superior or
more desirable than a forward approach?


Regards
Fabio