[Pce] Quick review of draft-hsd-pce-sr-p2mp-policy

Dhruv Dhody <dhruv.ietf@gmail.com> Mon, 15 July 2019 15:21 UTC

MIME-Version: 1.0
From: Dhruv Dhody <dhruv.ietf@gmail.com>
Date: Mon, 15 Jul 2019 20:50:19 +0530
Message-ID: <CAB75xn4pAkzkbvTQZspqHpCnUzibfQbrHfuEBVmYX+Qgvz1vXQ@mail.gmail.com>
To: draft-hsd-pce-sr-p2mp-policy@ietf.org
Cc: pce@ietf.org
Content-Type: text/plain; charset="UTF-8"
Archived-At: <https://mailarchive.ietf.org/arch/msg/pce/9b7Kh3I0wTgSFjyYbnFpLIF0vlw>
Subject: [Pce] Quick review of draft-hsd-pce-sr-p2mp-policy
Precedence: list

Hi Authors,

I did a quick review of your I-D. I hope this could help you better
prepare for your presentation slot in the PCE WG.

(1) I checked the I-D in spring [1] as well, but could not find a
figure that could show me how the SR forwarding would look like. I
would suggest you include a figure in your slides that shows label
stacks as the packet moves for -
- same Tree-SID used everywhere
- different Tree-SID at root/branch/leaf
- directly-connected v/s multi-hop v/s explicit-path between
replication points (branch nodes)

(2) Regarding the central control instruction, it would be good to
list, what information (in general terms) needs to be pushed towards -
- Ingress
- Replication Nodes
- Leaves
And hope this links up nicely with (1). You have some data in section
3.1 but it is not clear.

(3) The draft includes various details on -
- MVPN / PIM interactions
- Leaf discovery
- NBI at PCE
All of these should not be any different from P2MP RSVP-TE from the
PCE point of view, and the way you provide these details it feels like
it is some thing new needed to support SR P2MP. I would suggest that
the focus should be only on the *new* procedures and refer to existing
procedures as much as you can.

(4) Other considerations
- IMHO there is possibility to reuse lot more PCEP objects and TLVs
rather than defining them newly, this we can go over later if there is
interest and support from the WG
- IMHO regarding message encoding, you require a way to download
multiple SID-List at the root and branch nodes, to me Stateful P2MP
initiate message which allows Multiple ERO with each ERO having SR-ERO
subobject could achieve it, the things that are missing can be added
on this as a base. But lets focus on this part once we have clarity on
others.
- Global optimization in stateful PCE refers to optimization of
multiple LSPs, I think in section 3.2.2 you mean End-to-end
optimization?
- You need a new CCI object-type of the existing CCI object, rather
than a new PCEP object called SR-P2MP-CCI.

I hope you would focus on (1) and (2) during your presentation;
discussion on the list would be even better.

Thanks,
Dhruv

[1] https://datatracker.ietf.org/doc/draft-voyer-spring-sr-p2mp-policy/

[Pce] Quick review of draft-hsd-pce-sr-p2mp-policy Dhruv Dhody