PCE Working Group Meeting Tuesday, November 3, 2015; 5:10-6:40 PM PCE Chairs: Julien, Jon, JP (in absentia) Meeting Minutes by Dhruv Dhody Audio: http://ietf.org/audio/ietf94/ietf94-room501-20151103-1710.mp3 Meetecho: http://recs.conf.meetecho.com/Playout/watch.jsp?recording=IETF94_PCE&chapter=chapter_1 1. Introduction 1.1. Administrivia, Agenda Bashing (chairs, 5 min) * Welcome Jon as new co-chair! * Please try to get traction on the mailing list before requesting agenda. * Milestone update / re-charter in near future 1.2. WG Status (chairs, 20 min) [25/90] * Fatai Zhang: (For inter-layer) stable for long time, depends on GMPLS extn draft * Jonathan Hardwick: Just keep alive and we can go after GMPLS --- * Dan King: (for inter-area-as) Ready, (1) Need review (2) Should we include new work and make it becomes a living document * Jon: Not to have a living document * Dhruv Dhody volunteered to review --- * PCE-DISC-MIB and ENHANCED-ERROR about to be dropped, speak up if you care about this work * Please ask for early allocation only if your document is mature. --- * Dan (for H-PCE): There are 2 implementation (one on github) new version of document to be out soon, and check if it should be kept experimental. --- * Xian Zhang: (for GMPLS-initiated) Will update by next IETF. 2. Work in Progress 2.1. PCEP Extension for Association (Ina Minei/Xian Zhang, 10 min) [35/90] draft-minei-pce-association-group * Cyril Margaria: Value in this work, but not sure about process as the dependent work are non WG items * Ina: Chicken-and-egg problem * Julien: Regarding process if there is interest, we can adopt this, there is no issue! * Jeff Tantsura: It is important that this gets adopted, dependent draft will have implementation soon --- * Cyril: Association state is tied to PCE-IP address, it is too restrictive * Ina: But Association must note who created the association, how to handle association when that entity die? * Cyril: Fine with the encoding, but error procedures need to be clarified. --- * Lou Berger: Will respond when I see the mailing list that comments related to association encoding --- * Julien: 2.2. BGP-LS Extension for PCE Discovery (Jie Dong, 5 min) [40/90] draft-dong-pce-discovery-proto-bgp * Adrian Farrel: Your 2 use-case are good, but 3rd use-case is confusing? * Jie Dong: The PCC may not be a BGP speaker and in that case we can use IGP discovery. * Adrian Farrel: That is an 'importing' information of node that is not even in the IGP network. This is interesting. * Jie Dong: How is this importing done in IGP is out of scope but part of the solution. * Julien: The use-case/requirements and code-points needs to be clarify and improved. 2.3. Auto-Bandwidth (Rakesh Gandhi, 10 min) [50/90] draft-dhody-pce-stateful-pce-auto-bandwidth * Julien: Follow RFC7120 for early allocation, should be consensus and it is not stable yet. * Loa Anderson: Early allocation is based on stability of IANA allocation part and not the entire specification. * Julien: * Rakesh Gandhi: We need early allocation for all TLVs * Julien: Not sure if we have enough support. * Dhruv: Can we poll anyway? * Julien: * Jon: Please send to the list if you agree with work on the list with comments, silence on mailing list is harder to judge consensus. * Julien: Make sure that proposal will scale in large networks * Rakesh: Various scalability issues are handled in the current document. One criteria for adoption is to look into, if it is a problem that WG needs to work on. 2.4. Possible Follow-up on PCE-CC Proposal (Adrian Farrel, 15 min) [65/90] * Ina: Another function would be to use PCEP to get state from the network * Adrian: Yes, PCEP-LS is another such function * Julien: on agenda, will be discussed then --- * Lou: Where do you want to go with this with presentation? * Adrian: WG should make consensus decision about it * Lou: During re-organization, we decided that architecture should be done in TEAS. I support notion that not to do things ad-hoc, lets repeat discussion in TEAS WG * Adrian: Strike the WG in the slide, add IETF! --- * Ina: Thanks for bringing this up, it is better if there is a standardized way of doing this rather than doing proprietary stuff to fill these gaps --- * Danielle Ciccarelli: I agree that the 1st functionality is not needed for ACTN, but why do you think the 2nd is not needed? * Adrian: I dont think ACTN motivates these functions, ACTN might use these * Danielle: Agreed --- * Lou: In your opinion, are these function needed? * Adrian: The 2nd one is needed. The 1st one I remained to be convinced. In SDN model, I remained to be convinced that PCEP should be used as SBI, but I can see that it can be used as SBI. Deployment/implementations support will convince me. * Lou: From protocol perspective, is there zero work to do it? * Adrian: Yes, but is it sane to do this? --- * Sergio Belliotti: What is the intention for this presentation? * Adrian: No agenda, I am drawing attention to these draft, discuss the function and decide if we should have them. * Julien: There are two many use-case documents these days, we can sometimes aim for protocol specifications directly. * Lou: Send the functions to TEAS, protocol implementation in PCEP. --- * Chao Zhou(Cisco): Based on my talk with customers, the downloading speed of OF and NetConf is not meeting the customers requirements and perhaps PCEP is the answer. 3. I-Ds not Discussed 3.1. LSP Policies (Jeff Tantsura, 5 min) [70/90] draft-sivabalan-pce-policy-identifier * Dhruv: Real use-case, in current version you are not using ASSOCIATION object, there is another document about attribute association, we can work together and move towards common mechanism. * Jeff: Absolutely --- * Cyril: Is there a Yang model for it? * Jeff: There needs to be an abstracted view of policy and an pointer to it. * Robin: There are lot of IDs proposed, there is policy ID, association ID, in IDR there is path ID, redirection ID... Can we unify this via a common name with an ID/Type? * Jeff: Ack, we can discuss with them ---- * Sergio: Do you provide a mechanism to map the policy with the policy-ID? * Jeff: Policy is defined independently 3.2. RSVP-TE Local Protection (Cyril Margaria, 5 min) [75/90] draft-cbrt-pce-stateful-local-protection * Rakesh: Who computes the merge-point label? * Cyril: Will discuss with authors * Robin: This draft just creates association between primary and bypass, Label/ERO should all be done by PCC * Jon: More text on procedures are needed 3.3. PCEP Extension for TE Distribution (Dhruv Dhody, 10 min) [85/90] draft-dhodylee-pce-pcep-ls draft-kondreddy-pce-pcep-ls-sync-optimizations draft-wu-pce-pcep-ls-sr-extension * 3.4. PCE-Initiated IP Tunnel (Xia Chen, 5 min) [90/90] draft-chen-pce-pce-initiated-ip-tunnel * --Meeting adjourned--