IETF-92 IS-IS WG Minutes Time Slot (150m): Friday, March 27th, 2015 0900-1130 CDT Chairs: Chris Hopps, Hannes Gredler Scribe: Acee Lindem (acee@cisco.com) Presentation Notes From: R. Parameswaran * WG Status - Chris Hopps - See slides * YANG ISIS Data Model Stephane Litkowski - See slides - Presentation Notes: . Notion of node-tag. . Adding key-chain for authentication. . Introduction of fast re-reroute. . We just created global containers. These do not enable LFA, that happens at the LFA level. . Liitle bit of SR introduced to the ISIS model. . VRF centric vs. Protocol centric . Need feedback - No Discusion. * IS-IS Path Computation and Reservation Janos Farkas - See slides - Presentation Notes: . Explicit Trees . 802.1Qca is entering Sponsor Ballot. Will go for two iterations. . Requires new TLVs in ISIS, draft in ISIS and discussed here. . IS-IS draft synced with Qca 1.4 . Topology sub-TLV updated. . Bandwidth assignment sub-TLV . Timestamp sub-TLV added. . Added section for Symmetric paths. . Clarifications on MRT. . Description of multi-block GADAG . Leverages existing TLVs and algorithm . A few new sub-TLVs . Request adoption. - Discussion: Hannes: Would like to start the WG adoption call after IETF 92. Janos should initiate early allocation process for codepoints. Would hope to publish the document in less than a year as not to hold up the IEEE usage for SPB. * BIER support via ISIS Tony Przygienda - See slides - Presentation Notes: . Between -01 and -02 . Sub domains introduced, mapped to topology, sub-domain has a single tree. and single encap. . Any router not advertising enough labels to cover all BFERs are ignored. . Experimental idea - backwards compatibility bit. . TLV with min, max range. . Routers may fall back to min or run at the newest revision of the protocol. . SR uses algorithm in TLV versus tree-type here. . Other option is to have two separate TLVs, that's equivalent. . Even with identical TLV types, usually interoperability issues ae seen. - Discussion: Acee: Minor point - segment routing uses the term "algorithm" rather than "tree type". Tony: Will look at aligning. Hannes: First time versioning in sub-TLV. Tony: Need this level of granularity for use-case. Hannes: Used for atomicity? Tony: Yes - future hook for defining range of versions. Chris Hopps: Could you just use 2 TLVs. Chris Bowers: Experimental seems reasonable. Broadens the testing space with support of many versions. Tony: Same as using 2 TLVs to define range of versions. Chris Bowers: Still more permutations to test. Les Ginsberg: Version is a new parameter to track in codepoint allocation. Need to describe transitions between versions in the text. Tony: Yes - documenting transitions needed. It is experimental. This gives you more possibilities. * IS-IS BIER BFR-id Automatic Assignment Tony Przygienda - See slides - Presentation Notes: . BFR-id cannot be manually assigned in large domain. . Auto Assignment. . Centralized - single point of failure. . Distributed. . D-BFR elected and signal BFR-ids within sub-domains. . Will be backwards compatible. - Discussion: Acee: When you define sub-domain, you need to define what routers participate. Could you not define BFR-ID along with membership? Tony: This eliminates one configuration step. Define a sub-domain that maps to a service - PEs configure the service. * Advertising S-BFD Descriminators in IS-IS - Les Ginsberg - See slides Les Ginsberg - See slides - Presentation Notes: . One code point change. - Discussion: Will ask for WG last call when BFD working group last calls the base S-BFD document. * IS-IS Route Preference for Extended IP and IPv6 Reachability Les Ginsberg - See slides - Presentation Notes: . Waiting for prefix attributes draft (see next presentation) * IS-IS Prefix Attributes for Extended IP and IPv6 Reachability Les Ginsberg - See slides - Presentation Notes: . Competing proposals, reached agreement - will use prefix attributes draft. - Discussion: Needs WG adoption in support of other WG drafts Chris Hopps: Anyone object it being a WG document? Room: No objections. Chris Hopps: Will confirm on the list. * Advertising Encapsulation Capability using IS-IS Uma Chunduri - See slides - Presentation Notes: . Incremental deployment of new technologies. . New TLV in ISIS to carry encap type. - Discussion: Ahmed Bashandy: Problem not specific to SR or R-LFA. This is about a 15 or 20 year old problem. It has been solved without this extension. I see no need to solve this problem. Xiaohu Xu: There is already a routing extension in BGP. This is simply the IS-IS equivalent. Les: Could use prefix admin tags and it has the advantage of tying the endpoint to the tunnel. Uma: Tag today is only used for leaking between levels. This approach is more straight forward. Les: For the record, I don't think this is difficult (i.e., using prefix admin tags). Xiaohu: Has noted other ways but this is better. Ahmed: Trying to solve 15 year old problem without any complaints. Uma: Ingress node can dynamically form tunnel for SR, BIER, and other usecases. Hannes: This is the formal equivalent of the tunnel SAFI in BGP. So this is not an unsolved problem - just moving solution from BGP. Xiaohu: BIER has a usecase where you need to traverse a non-BIER node. Tony: Can use BIER sub-TLV to identify non-BIER and tunneling. Ahmed: IS-IS config required - admin tags can be mapped to tunnel types. Hannes: Same argument can be made for any feature or routing itself. There are requirements for auto-tunneling and this is a straight forward extension. * IS-IS Link Overload Shraddha Hegde - See slides - Presentation Notes: . Corresponding draft presented in OSPF as well. . Explained need with topology. . Link overload bit in the link attributes sub-TLV. . Metric increased to MAXMETRIC. - Discussion: Les: Why would you put this bit in the LSP advertisement other rather than hellos. Shraddha: Uses existing TLV encodings. Les: This forces you to advertise all parallel links where this is not required today. Les: Did you look at the reserse-metric draft? Shraddha: Just learned about it. Les: Believes if this is required, the reverse-metric draft is the right way to do this. Uma: Small correction referencing DR/BDR which is OSPF terminology. This changes the metric on the fly. Pushpais: Agrees with comment on reverse metric. Is there an OSPF reverse metric? Acee: No reverse metric but we have a 2-part metric draft. Hannes: Should use the same mechanism in both IGPs? Acee: In OSPF, we will discuss whether this mechanism is needed in the first place. Les: Mechanisms are the same but encodings can be different. Still thinks hello rather than LSP/LSA is the best way to solve this problem. * IS-IS Topology-Transparent Zone - Huaimo Chen - See slides - Presentation Notes: . Motivation - growth of network topologies. - service disruption and time consuming. - complex for End-to-End Soln. . Links, routers inside TTZ are not advertised to routers outside TTZ. . TTZ TLV - TTZ Id, and sub-TLVs. No sub-TLV means all links on router are abstracted out (and router itself as well). . TTZ IS Neighbor sub-TLV. E bit means edge router. . Only tagged links are advertized outside the TTZ zone. . TTZ TLV with TTZ Id in Hellos. . For migration. . Config TTZ on all routers. . confgigure TTZ on edge router to allow propagation outside the TTZ zone . 1 command to edge router to migrate cloud to TTZ. . Request feedback, adoption. - Discussion: Les: It is possible with the base protocol to merge and split areas non-disruptively. Mulitple area addresses can be configured for the same area allowing a smooth migration. Why invent a new extension for this requirement? Huaimo: No need to flap links to change areas in IS-IS? Les: Migration techniques are straight forward and there is no need to flap links. Chris Hopps: Document the migration steps on the list. Renwei: TTZ can solve much more than changing area topology. Can migrate networks between routers and can reduce the size of the IS-IS network topology. Hannes: Can use multi-instance to model a TTZ. Already tooling in the IS-IS tool box. Would like to see this used rather than a new mechanism. Huaimo: Doesn't believe this solves all the same problems. Chris Hopps: Please take this discussion to the list. Chris Hopps: If no more comments, meeting adjorned. We will try harder not to get a Friday IS-IS meeting in Prague.