SPRING WG - Source Packet Routing in Networking Monday, July 16, 2018 13:30-15:30 Monday Afternoon session I Room: Duluth ====================================================== Chairs: Bruno Decraene Rob Shakir o Administrativia Chairs - Note Well - Scribe - Blue Sheets - Document Status 15 minutes 13:30 https://www.youtube.com/watch?v=rJKC14C4Gj4&t=30 Rob: new charter need to focus on deployability of SR, the operational aspects of deploying SR. Rob: due to lack of tight agenda, the SR-TE-policy will be discussed in IDR WG, multicast draft in PIM. Jeff T: Multicast draft will also be discussed in RTGwg o SR YANG model draft-ietf-spring-sr-yang-09.txt slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=521 Yingzhen 5 minutes 13:45 It has been 2 years since we first presented it. It is very stable now. Main point: SR Global Block for entire network configuration; Maxi SID; notification Define a Transport Type at the very beginning. We think the feature wise is complete. If you have any concern, please voice. we think it is ready for WGLC discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=757 Bruno: we will Yang doctor review, then start the WGLC. [D Jain Cisco] will common types for SR be defined in this module or elsewhere? [A] Two models are defined within this draft, one of which defines the common types. [Jeff T] the initial idea to have base here and reference other YANG models. There are many others being developed so far. [A] Encourage authors of other YANG models to reference this one if they need common types for SR. (Chairs note - t=0, 13:45) o SR policy draft-ietf-spring-segment-routing-policy slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=900 Ketan 10 minutes 13:50 Major change: the document is re-organized, focus on Core SR policy. Others are moved to other documents. discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=1190 [Andrew Stone Nokia] There is a BGP extension for SR-TE policy. Any extension for PCEP? [Ketan] Partially covered in BGP [Dhruv Huawei] we have a document covered in PCE WG. I can send the link [Harish Sitaraman Juniper] There are symbolic names for SR-TE policy. Any extension for the BGP SR-TE draft? [Ketan] the latest version of BGP has that included [Harish] the draft said the name must not be used as policy identification. [Ketan] Information model has an index. Name is not unique. [Sue Hares] are you going to harmonize BGP and PCEP extension? we should follow up afterwards. (note: 8:10) draft-filsfils-spring-sr-policy-considerations slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=1399 Objective of the document is to document implementation aspects and which WG should cover what aspects Asking for WG adoption. discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=1740 [Rob] Few challenges with adopting this draft. A lot are implementations specific e.g. splitting functionality across software modules, some of which may be implementation specific. What constitute Traffic Steering, what is traffic policy. Given that this document discuss path computation we need discussion with TEAS. We need to go through these discussions before we can kick of the adoption call. [A] Some part are conceptual, they don't have to be implemented as written. [Rob] May be editorial, some editorial restructuring may help. (note 16:50) o NSH and Segment Routing Integration for Service Function Chaining (SFC) slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=1914 draft-guichard-sfc-nsh-sr Jim Guichard 10 minutes 14:00 The purpose of the draft to clarify the confusion people think that SFC might be competing with SR. The truth is NOT. SFC is complimentary to SPRING. SFC NSH Header can work with any transport technology. NSH can integrate with SR for identifying flow. Using NSH to identify which context. discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=2330 [Andrew D Nokia] we would like people to look at this carefully. SFC was done independently from transport layer. The more you put service into transport layer, the messier it gets. I support this separation proposed by the draft. (25:03) o Service Programming with Segment Routing draft-xuclad-spring-sr-service-programming-00 François Clad 5 minutes 14:10 slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=2415 Service are integrated with SR policies. discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=2766 Bruno: are you going to utilize existing way to carry the metadata or specify a new way to carry metadata? Francois: we plan to create a new metadata TLV. Joe Harlpern: you are avoiding interoperability discussion. we have SFC over MPLS, SFC over IPv6. What you are doing is reinventing entire new thing. SR is not new to invent completely new Francois: NSH needs a lot of state in the network Joe Halpern: can you name a few "state"? yes there are trade-off, but we can't reinvent a new TLV for every corner case Jim Guichard: Let's use the technology that we have developed, rather than reinvent everything. Let's collaborate, lets' not reinvent it. Zafar Ali: if we are afraid of touching anything new, then we can't do anything. What is wrong to improve to remove layers in the network? Andrew D Nokia: please don't make statement on STATE. NSH is for services, SR for transport. It's not because we can make it simpler for a simpler use case, that we need to develop a new solution. Whole picture would be more complex both for vendors and operators. Rob: cutting the line. [Jabber: Greg] how you envision separation of Transport OAM and Service OAM? [A] Would use the same solution to check the transport and the service. There is another draft on OAM, I will check if this is covered. Rob, Bruno: continue the talk offline. (40:00) o SRv6 Network Programming draft-filsfils-spring-srv6-network-programming-05 Pablo Camarillo 10 minutes 14:15 Slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=3340 Discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=3720 Ron Bonica (Juniper): why is it necessary to re-encode VPN MPLS label? why need to move the VPN info into the IP destination address? Darren (Cisco): it is about the IPv6 SR data plane vs MPLS data plane. to be answered in text [?] Joel Halpern: more confused about the Transit SIDs. [A] disagreement with your reading of the draft Uma: related draft-ietf-dmm-srv6-mobile-uplane. SRv6 can be used for underlay, when you remove TE tunnel. Draft requires LTE network to rely on IPv6 addresses while currnet deployement may use MPLS or IPv4 transport. GTP separation is not good. o Node Protection for SR-TE Paths draft-hegde-spring-node-protection-for-sr-te-paths Chris Bowers 10 minutes 14:25 Slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=4090 Remotely presented. Discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=4900 Bruno: which WG (RTGWG or SPRING)? Chris: more appropriate in SPRING. The need to get to SR-TE path Stewart: do you need to dynamically compute the path? Chris: Protection is computed ahead of time Stewart: it conflicts with the SPRING principle of intermediate nodes not knowing the state. Chris: you have to violate somewhat for the second label (1h07) o Supporting Flexible Algorithm Prefix SIDs in LSP Ping/Traceroute draft-iqbal-spring-mpls-ping-algo-00 Faisal 8 minutes 14:35 Slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=5020 no questions (1h12) o Segment routing for SD-WAN paths over hybrid networks draft-dunbar-sr-sdwan-over-hybrid-networks-01 Linda Dunbar 6 minutes 14:43 Slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=5240 Discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=5913 Jeff Tantsura: there is management element in there. need to talk to Ignas to see which WG needs to be discussed [Insufficient time for Robin's question - to the list] o SR For SDWAN: VPN with Underlay SLA draft-dukes-spring-sr-for-sdwan-00 Darren Dukes 6 minutes 14:49 Slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=5948 Discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=6200 Darren: Will discuss with Linda on how to integrate the two drafts on SD-WAN tomorrow afternoon (Tuesday at 3:30pm at IETF registration) Daniel (Bell Canada): some similarities with the previous draft, I guess you will need to collaborate both draft together. Darren: Absolutely. o Operations, Administration, and Maintenance (OAM) in Segment Routing Networks with IPv6 Data plane (SRv6) draft-ali-spring-srv6-oam-00 Zafar Ali 10 minutes 14:55 Slides: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=6256 Discussion: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=6735 Jabbar (Greg): purpose and how to use the time-stamp. Zafar: timestamp is implementation dependent; as soon as possible on the ingress line card. Ron Bonica: (SRv6 Overlay Traceroute slide) This page includes other routing header types. It might be useful to make it generatic by moving the O-bit out of the SRH to a destination option. A: easier to get the flag from the same header; behavior specific to the SR SID. o Segment Routing Traffic Accounting Counters draft-filsfils-spring-sr-traffic-counters-00 Zafar Ali 10 minutes 15:05 Bruno: Lack of time for this presentation Please summarize the 2 slides on the mailing list Zafar: We like to ask for WG adoption Bruno: ok, request can be also be part of the email o Performance Measurement in Segment Routing Networks SR-MPLS: draft-gandhi-spring-sr-mpls-pm-01 SRv6: draft-ali-spring-srv6-pm-02 UDP Path for In-band Performance Measurement draft-gandhi-spring-udp-pm-01 Rakesh Gandhi / Zafar Ali 10 minutes 15:15 Presentation: https://www.youtube.com/watch?v=rJKC14C4Gj4&t=7027 No time for questions. Speaker Shuffling Time 5 minutes Total 120 minutes 15:30