PCE Notes Tuesday November 18, 2008, Afternoon session 1:00PM- 1) Administrivia (Chairs) - 5 minutes [5] No comments 2) WG status (Chairs) 15 minutes [20] No comments 3) RSVP Extensions for Path Key Support (Adrian - 5mn) [25] draft-ietf-ccamp-path-key-ero-02.txt Lou Berger: question raised in CCAMP about the number of bits used to encode a path key. Adrian: Are 16 bits enough for the path key ID in the pay key object? How many active IDs will a PCE have at any time? How many LSPs using path keys? How long are they active? Could increase this to 32 bits if folks are worried? A PCE could use a second PCE ID to increase the available path keys… Take discussion to the list. 4) A set of monitoring tools for Path Computation Element based Architecture (JP - 5mn) [30] draft-ietf-pce-monitoring-03.txt JP> Drat has been stable for a while - Update on the latest version addressing editorial issues. Need to find volunteers to read this before last call. 5) Vendor Constraints (Adrian - 5mn) [35] draft-farrel-pce-vendor-constraints-02.txt Quick review of issue and solution suggested in the draft. Enterprise IDs. Poll for WG status. => looks like consensus. Will take to the list. Don O'Connor: What are the bounds on this? Competing vendors? Cooperating Vendors? Adrian: No bounds except the allocation of an Enterprise ID. JP: Need to add text to say that if a new extension is planned to become a standard, it should come with its own Internet-Draft and request new code points. It should not use the vendor specific constraints. 6) Update on P2MP PCEP extensions (Quintin Zhao - 10mn) [45] draft-ietf-pce-pcep-p2mp-extensions-01.txt JP: What if different PCs request leaves for the same tree? Will provide scenarios to explain what he means. JP: Have you worked with service providers on issues such as how many leaves and how much one would need to do defragmentation? Size of the tree? Any service providers williing to share data? Adrian: Note that France Telecom is a co-author of this work. 7) PCEP Requirements and Extensions for WSON (Young Lee - 5mn) [50] draft-lee-pce-wson-routing-wavelength-03.txt Lou Berger: Have you thought about separating this into two documents one on requirements and one on mechanisms? Young: Good idea. Lou: What of this is general to GMPLS and what is specific to WSON? Don O'Connor: Support separating general GMPLS work from WSON work. In particular the separation of computations for different pieces of the problem. Igor Bryskin: Concern that some models in this I-D cannot be made to work. JP: Delay making this as a working group document until CCAMP WSON work has stablised. 8) Alternative Approaches to Traffic Engineering Database Creation and Maintenance for Path Computation Elements (Greg Bernstein - 10mn) [60] draft-lee-pce-ted-alternatives-00.txt JP: Is this the right time for this work? Perhaps as we now have the core pieces of PCE in place. Igor Bryskin: Strongly believes that this mechanism will create an up-to-date TED faster and result in a more accurate computation (compared to flooding). Adrian: We clearly need to trade off scaling implications and flooding speed. JP: The order of magnitude of improvement is small. Igor: Likes the ideas in this I-D. Three problems come to mind: 1. The draft impies the use of PCE for all computations. I would like to split the info with some in the IGP and some sent to the PCE. Greg: Yes, we could do that Igor: 2. How do we handle stale information, etc. The solution has to complement the way the IGP handles stale information. 3. What will the traffic congestion be like around the PCE? JP: TCP will scale OK around the PCE. We know that TCP can support thousands of simultaneous connections. Kam Lam: If there is more than one PCE in the domain (for example, for robustness) we must consider how to keep the TEDs synchronized. Greg: Synchronization between PCEs for TED doesn’t necessarily need to be standardized. 9) PCEP Extensions for L3VPNs (Kenji - 5mn) [65] draft-kumaki-murai-pce-pcep-extension-l3vpn-01.txt JP: Did you look at the signaling issues? Are there any drafts in other working groups addressing the signaling? You need to consider several points: • Signaling an LSP between two CEs… • Issues with RSVP-TE (IPv4 and VPNv4 addresses) • Issue with fast reroute for Node Protection of the PE Adrian: In Dublin, we suggested aligning with the draft on VPN requirements? Have you done that? Kenji: We have similar requirements Adrian: If we know that both drafts are moving in a similar direction it is easier to take as a WG document. Please try to harmonize with the VPN requirements draft and where possible, generalize your solution. JP: The chairs will drop you an e-mail to explain the issues. 10) Discovering PCEs in other ASes: (Kenji - 5mn) [70] draft-kumaki-pce-bgp-disco-attribute-02.txt Adrian: Who has read the draft ? very few. JP: Have you talked with the L3VPN working group? Kenji: Yes, some feedback. JP: Need to do an overall architecture for the operation of L3VPNs with PCE covering both these drafts and both working groups. Need to have a look at the overall problem of signaling and fast reroute including CE-CE tunnels. To proceed further with this, we suggest you go back to L3VPN and work through the other issues. 11) PCE communication protocol(PCEP) MIB (Kiran – 10mn) [80] draft-kkoushik-pce-pcep-mib-02.txt JP: Its part of the charter we should make this a working group document. Adrian: The read/write thing. What would there be that would be writable? What about using the MIB for requests and responses?