Hi JL,
In Vancouver you said you had some thoughts and issues with the P2MP PCEP
drafts draft-yasukawa-pce-p2mp-app-01.txt and
draft-yasukawa-pce-p2mp-req-04.txt.
The main issue that you raised at the microphone was that you felt that the
I-Ds were too strong in their assertion that P2MP computation would never be
done at a head-end LSR.
Looking back at the I-Ds, I see that the requirements draft makes no claims
about the applicability of the PCE architecture to P2MP LSP computation: it
devolves this responsibility to the applicability I-D.
The applicability I-D does discuss the potentially greater computational
load for P2MP LSPs compared to P2P LSPs. It states:
Roughly speaking, the load to compute a least-cost-to-leaf tree is
the same as the cost to compute a single optimal path to each leaf in
turn. The load to compute a Steiner tree is approximately an order of
magnitude greater.
From what you said in Vancouver, you are suggesting that approximations to
Steiner may be achieved in roughly the same order of magnitude as a
least-cost-to-leaf tree. Would be enough to you if we added to this
paragraph to say...
...although powerful heuristics exist that may make it possible
to approximate a Steiner tree in the same computational time
as a least-cost-to-leaf tree.
Were there any other issues you wanted to raise?
Cheers,
Adrian
_______________________________________________
Pce mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pce
