Minutes DetNet Agenda IETF98 (Chicago) > Version: Mar 23, 2017 > > Session 1: > Monday, March 27, 2017 (CDT) > 13:00-15:00 Monday Afternoon session I > Zurich G > > Etherpad: http://etherpad.tools.ietf.org:9000/p/notes-ietf-98-detnet?useMonospaceFont=true > Meetecho: http://www.meetecho.com/ietf98/detnet > Audio stream: http://ietf98streaming.dnsalias.net/ietf/ietf988.m3u > Jabber: xmpp:detnet@jabber.ietf.org?join > > Available post session: > Recording: https://www.ietf.org/audio/ietf98/ietf98-zurichg-20170327-1300.mp3 > YouTube: https://www.youtube.com/watch?v=HW-zShuDHh4 > Note takers: Jouni Ethan Lou > # Start Duration Information > 0 13:00 15 Title: Intro, WG Status, Draft Status > Presenter: Chairs Lou Berger: Where do we stand on WG documents not on the agenda? Authors? Norm Finn: Architecture: The authors are happy with the document Pascal Thubert: The authors are happy, but there have been some comment on the list regarding text from the data plane documents that have not been addresses. Also need to ensure that the draft is aligned with the latest data plane text. Lou Berger: The chairs and DT lead can get with you to try to close on this document Ethan Grossman: use cases ready for WG LC (current draft 11 will expire in 10 days - needs refresh) Jouni Korhonen: for the data plane alternatives, one more spin to add the final conclusion and then ready. Pascal Thubert: problem statement good to go. What do the chairs want to do? Lou Berger: if the document can be progressed without much distraction of the WG, would be happy to see it published. (service model text distributed into other documents.) > 1 13:15 30 Title: DetNet Data Plane Design Team > (+00:13) Presenter: Jouni Korhonen > Draft: https://tools.ietf.org/html/draft-dt-detnet-dp-sol-00 Stewart Bryant: Open item for DP: IP different design from MPLS. Most IP based solutions use L2TP which doesn?t use MS-PW. RFC3985 has a fragmented design for IP over MPLS, but this isn?t how most have implemented PW over IP. Jouni Korhonen: DT took the route of MPLS over IP. It follows basically what has been described in rfc3985 and TDMoIP/SAToP (RFCs 5087/4553). Stewart Bryant: concerns regarding sequence number handling (from PREF) in a multi-port systems. This is a hard issue. Jouni Korhonen: DT looked into this and determined (for now) that is a device design has multiple linecards (or equivalent architecture) where flows end up under processing of different entities, then the solution implementation has to solve the issue somehow. Stewart Bryant: We looked into this in the context of packet loss measurement and the only solution we?ve come up with is alternate marking, and no one has figured out a solution at these sorts of speeds. Andy Mallis (cochair of PALS WG): Sees two drafts to bring into PALS: 1 Detnet PW, including definition of exit PE behavior and duplicate elimination; second is update to 6071 MS-PW to include new definition of S-PE and T-PE behavior. Lou Berger: We?ll have to discuss this with the AD and Chairs. Deborah Brungard (AD): Common to move work between groups as appropriate Norm Finn: The common solution being taken to solve the elimination problem is to look at it at the output port. Norm Finn: Typically, in systems that make use of elimination the elimination function is implemented in the output port. That mitigates some of the issues. Lou Berger: How many have read the draft -- a reasonable number How many think is a reasonable document to adopt in *a* WG (pending discussion with PALS) - a reasonable number Including PALS Chair Andy Mallis: Not all goes to PALS, just changes to PWs and changes to MS-PWs. Stewart Bryant: Need to discuss with L2TP WG Lou Berger: L2TP isn?t currently being discussed in the WG Stewart Bryant: So the WG is going to follow the path that has never been built Lou Berger: Keep in mind this is the start of the WG activity, not finalizing a solution. Any decision taken here will be confirmed on the list. Eric Gray: We don't make decisions in the room we make it on the list Lou Berger: Agreed > 2 13:45 20 Title: DetNet Flow Information Model Based on TSN > (+00:43) Presenter: Bal?zs Varga > Draft: https://tools.ietf.org/html/draft-farkas-detnet-flow-information-model-00 Kiran ???: Are you looking at IOT, where there is no IP or MAC addresses? Bal?zs Varga: This hasn?t been discussed. Lou Berger: What about alignment with data plane solution document Bal?zs Varga: Isn?t there yet Lou Berger: PW over IP - assigning resources for QOS is understood in MPLS but not for IP. Bal?zs Varga: There is some part in the service model draft. This is central to the discussion. Yuanlong: Perhaps consider performance metrics end to end for flows? Bal?zs Varga: This is worth additional discussion. Yuanlong: There is an issue of setup latency/availability. What network provides may be diff't than requested? Bal?zs Varga: Parameters in info model e.g. latency. And need to reflect results of service request in response. Yuanlong: There was another flow info model presented in the last meeting Lou Berger: It would be good to see how the two different drafts can be reconciled/combined Bal?zs Varga: There are different models in the drafts, we based ours on the IEEE, but this is something to sort out. Yuanlong: Agreed. > 3 14:05 20 Title: DetNet Security Considerations > (+1:02) Presenter: Tal Mizrahi > Draft: https://tools.ietf.org/html/draft-sdt-detnet-security-00 Norm Finn: TSN reduces attack surface thanks to extensive protections against misbehaving flows, limits damage. Hard to tell between failures and attacks, but due to protection inherent, is hard to to mount DOS attack. Stewart Bryant: Are these problems tractable? I.e. dynamics rather than data contents. (apart from protecting edges) Tal Mizrahi: Looking at threats now, not yet solutions. Norm Finn: We do have time reference, so know when things are to happen or not. Packet must be at right time on right interface with right packet, otherwise don't trust it - will discard and shut down the operation. Isn't that the goal of the attacker, to cause shut down of operation? No, there is also a correct source of the data, so if inhibit attacking source, that doesn't cause damage to real source and thus operation of the system. Stewart Bryant: Work of security for time group is easier than for DetNet? Norm Finn: A lot of work on this in automotive, but don't have access to that work. Lou Berger: There is also interesting work in the IEEE Lou Berger: Is this (security) work important? (a good number) How many have read document (a reasonable number) How many think document is sufficiently mature for adoption (few) Will revisit adoption after next version. > 4 14:25 15 Title: IEEE TSN and IETF DetNet Issues > (+1:21) Presenter: Norm Finn > Draft: Norm Finn: Time that packets spend in queues dominates latency (could be in router or bridge, doesn't matter) Queues are associated with ports, control port parameters with YANG model. Eric Gray: Bridging TSN islands doesn't help end to end. ... (?) Norm Finn: Let IEEE define Q models, we can use them. Lou Berger: Yang models consistent with our scope, proposal (on slide 6) not a stretch. Control plane extensions are outside scope of our charter. YANG is in scope. Okay to keep in mind issues we may introduce for the control plane is also okay. For actual control plane extensions, we can discuss (charter) once other document have been delivered. George Swallow: Can we still work on control plane requirements? Lou Berger: Indirectly via flow information model and YANG to control detnet flows. Direct extensions require discussion. George Swallow: Control is complicated, good if in same docs or at least same people working on it, not thrown over the fence. Deborah Brungard: (As AD) Control work can be coordinated with the appropriate group. Subir Das: We?ve already had some related discussion would be good to capture in a document. Lou Berger: Need to focus on core deliverables, but fine to note impact where it is identified. George Swallow: it would be good to ensure there is alignment in this WG and not use other WGs for DetNet discussions. Norm Finn: regarding cooperation between 802.1TSN. IEEE is now doing the YANG models for 802.1. Available in github > 5 14:40 20 Title: DetNet Backhaul Networks > (+1:38) Presenter: Lun Shao > Draft: https://tools.ietf.org/html/draft-wang-detnet-joint-scheduling-00 > https://tools.ietf.org/html/draft-wang-detnet-backhaul-architecture-00 Pascal Thubert: Concern that doc focuses on WAI-PA, which is one of multiple standards for this and we have no good access to them. Perhaps 6TiSCH would be a better primary focus for an IETF draft. Lou Berger: Keep in mind that not doing control plane, only data plane now. Our interface to control is via YANG. There is work going on in TEAS WG in the context of ACTN which may be aligned with what you are discussing, maybe look at trying to coordinate with that work. Norm Finn: Is this a new use case Lou Berger: The use case authors said this isn?t a new use case but rather a way of implementing a use case Eric Gray - Factory automation with centralized control is a new use case. Lou Berger: And we already have that use case documented, this seems like ways of supporting a use case. > Adjourn 15:00 See you in Prague.