PCE Working Group Meeting - Tuesday, March 20, 2018; 13:30-15:30 Afternoon session I o Slides: https://datatracker.ietf.org/meeting/101/session/pce o Meetecho: http://www.meetecho.com/ietf101/pce o Chairs: Jonathan Hardwick, Julien Meuric (remote) o Secretery: Dhruv Dhody 1. Introduction 1.1. Administrivia, Agenda Bashing (chairs, 5 min) - 1.2. WG Status (chairs, 17 min) [22/120] Dhruv Dhody (for draft-ietf-pce-inter-area-as-applicability): Dan King is the editor, but one decision that is pending is about including Stateful PCE details in this I-D. Jonathan Hardwick: Better to include stateful PCE details, as that would be bit obsolete. Dhruv (for draft-ietf-pce-pcep-yang): update with new examples and clarification on its use, and now ready for YANG doctor review. Haomian Zheng: (for draft-ietf-pce-enhanced-errors): updated draft with new error types, and reference up-to-date. Need confirmation with WG on the direction. Dhruv (for draft-ietf-pce-stateful-pce-lsp-scheduling): made an update before this IETF and aligned to the published TEAS RFC, more reviews are requested. Dhruv (question on segment routing draft): We have PCEP segment routing extension (ready to sent to IESG) and then there is SR TE policy (via BGP), we have never discussed how they interact or work together; and there has been confusion on this while discussing with customers. Jon: Please send a pointer. 2. Other WGs I-Ds 2.1. Flagging MD5 out as an authentication mechanism (Stewart Bryant, 3 min) [25/120] draft-nslag-mpls-deprecate-md5-01 Jon: If you care about this, please go to MPLS WG session. 2.2. BIER-TE (Toerless Eckert, 15 min) [40/120] draft-eckert-teas-bier-te-framework-00 Jon: Thank you for bringing this to the PCE WG, we have worked on multicast for PCE, so it would be good to have further discussion as well as watch the progress in TEAS WG. 3. WG I-Ds 3.1. Stateful PCE for GMPLS (Young Lee, 10 min) [50/120] draft-ietf-pce-pcep-stateful-pce-gmpls-08 Jon: Time to get some reviews, as we prepare for WG LC 3.2. Update on Association Drafts (Dhruv Dhody, 10 min) [60/120] Various Drafts Jon: I will do my shepherd review, once done we could start looking at other WG I-D. Time to poll adoption for protection draft. Open to discuss the order of other related drafts if there any dependency or we could approach in the order they arrive. This is a key piece and we formalize the idea of associating the LSPs and thus we need to make sure that the base association group draft is bang on! 3.3. Stateful H-PCE (Dhruv Dhody, 10 min) [70/120] draft-ietf-pce-stateful-hpce-04 draft-ietf-pce-hierarchy-extensions-04 Jon: Good to see H-PCE extn revived, and we could last call this now. Will forward these works in that order. 3.4. ACTN Applicability (Young Lee, 5 min) [75/120] draft-ietf-pce-applicability-actn-05 Jon: Who has read? (about 10) Jon: The document is stable and would be put in the queue. We are looking for shepherd if someone can. 4. New I-Ds 4.1. Relax Constraints in Stateful PCE (Stephane Litkowski, 10 min) [85/120] draft-dhody-pce-stateful-pce-optional-00 Adrian Farrel: Absolutely a problem to solve. Could you end up with a mechanism to trade options in a flexible/horrible way by ranking them. Stephane: This exist for stateless already. Adrian: This could be worse. Jon: (as an individual) A valid problem, and a reasonable approach, we may consider to make this more flexible in future. Question: Relationship with the relax-constraint draft? Stephene: Relax-Constraints TLV is removed from the diversity association draft and we choose this generic method (using existing flags). Jon: Prefer to keep it simple. 4.2. PCEP extensions for SR-TP (Xiong Quan, 10 min) [95/120] draft-xiong-pce-pcep-extension-sr-tp-00 Adrian: You have something good here! But what actually is SR-TP (you just have SR LSP)? Xiong Quan: A new network where SR technology is used in MPLS-TP (SR- TP in short) Adrian: Are you after the Central Control aspects and the PCE approach to program the head-end. SR-TP is marketing. Does the forward and reverse LSP have to be the same technology i.e. SR one way, conventional LSP the other? Pointing to another tunnel and not the type of tunnel would be useful. Dhruv: Can this be done with existing SR-ERO and the bi-dir association object? What is the value of path label sub-object and why is it needed? Himanshu Shah: MPLS-TP is well known, as Dhruv said, is it possible to implement this with what we already have by configuration on both ends? Xiong: SR is uni-directional, for bi-directional SR we need to bind the two directions and path label is required for SR network. Xin Wan: (slide 2) We notice a "controller" in the problem statement draft which implements SR-TP, which may be different with the PCE function, you may need to specify the difference. Stewart Bryant: The reason for the additional path label is not clear, as you have pseudowire over MPLS-TP and pseudowire label can do this! Jon: Explain why is the additional "path-label" needed? Stewart: Some mechanism is needed for sure, because this is SR and label is lost till the packet reaches the end, but can pseudowire label can play that role. 5. Previously Discussed Topics 5.1. Residual Bandwidth (Daniele Ceccarelli, 5 min) [100/120] draft-lazzeri-pce-residual-bw-01 Jon: How many have read the latest version (about half dozen). Jon: How many think this is a useful to work on (4-5). Jon: This version has clearer technical requirements, there is mild interest, will confirm on the list. 5.2. PCEP for Native IP (Aijun Wang, 5 min) [105/120] draft-wang-pce-pcep-extension-native-ip-01 Jon: This is an application of PCE that is a bit outside its usual use, in an IP only network. TEAS WG has adopted this work. Jon: Who is aware of this work and in the TEAS? (around 10) Jon: Who would like to see this adopted in PCE? (around 8) Jon: Who would not like this being worked on? (no one) Jon: Will confirm on the mailing list! 6. Update on SDN discussion (Chairs, 15 mins) [120/120] Adrian: I find this is hard, I can see reason for/against. When BGP- LS was proposed, we were not sure if that was a good idea. Get feedback from operations community or OPS AD. Jon: Good Idea. Robin Li: IGP is really not a good option because of multi-hop (need to setup tunnel). Jon: We need to get a good idea on why BGP-LS and Netconf is not suitable? Robin: Netconf's performance is not good enough, so the only candidate left is BGP-LS. Regarding inter-op we need to consider controller as well as device. Some operator are scared of BGP. Jon: BGP-LS is not quite BGP. You don't need full blown BGP. Haomian: Good summary. Presence of other ways to do this should not have a technical impact. The Link State is a important part of preparing PCE for path computation function. The PCEP-LS does not replace any existing technology, it is just another option. There is another draft in CCAMP that discuss the interaction between various protocols in different scenarios, but PCEP is an important piece. Young: For optical vendor, BGP-LS is not possible (no one implements that). There is a study on large scale network where PCEP-LS sending state northbound leads to faster convergence esp with lot of optical data (making PCEP-LS faster). Himanshu: Neutral to this discussion, if we already have PCEP connection, reuse of PCEP is a good idea (without implementing another protocol - IGP, BGP etc). That could be in favor of PCEP-LS! Jon: We will continue on the list!