------------- Monday, March 19, 2018 (GMT) 15:50-17:20 Monday Afternoon session II -3E Palace C tsv alto Application-Layer Traffic Optimization 13 participants Meetecho 20 parts in room Chair: Jan Seedorf Note takers: Sabine, Richard SLIDES: https://datatracker.ietf.org/meeting/materials/ => TSV Transport Area / alto ACRONYMS: BT Brian Trammel DC Dawn Chen JS Jan Seedorf JP Jon Peterson PANRG Path Aware Networking RG PV Path vector RY Richard Yang SL Shawn Lin TAPS Transport Services WG UP Unified Properties VG Vijay Gurbani ---------- Status of WG: by Jan Seedorf WG Chair https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-chair-slides-ietf-101-00 We need to focus on WG items. - RFC8189 on Multi-Cost published, congrats. - Cross Domain Server Discovery: should move. Chairs think ready for WGLC - ALTO Calendar WGLC is over and the draft will move to the ADs. - ALTO Performance Metrics: is under review from other expert upon Mirja's recommendation - CDNI FCI with ALTO: needs more work to get ready for WGLC - SSE: is close to WGLC - Path Vector and Unified Properties: move now as a bundle. Cell addresses will be added later on. ---------- "Path Aware Networking Research Group" by Brian Trammel PANRG chair https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-overlap-with-panrg-00 Brian exposes overlaps between PANRG and ALTO. PANRG meeting takes place on Tuesday. - Why PANRG? PAN expands multi-path transport beyond multi-connectivity; looks at alternatives on the current not path-aware Internet architecture for routing and trustful routing information; want to experiment cooperative signalling, how to tell what you want from the network. - Relation to ALTO: ALTO addresses 2 top PANRG questions: (1) "how define and represent path properties?" where path selection properties are being proposed in the TAPS architecture. (2) "How give EPs access to trustworthy information about path properties?" - PANRG also has science fiction questions (see slides) Discussion - Vijay on meetecho: are there real measurements on how devices used in multi-paths or academic papers? BT: question for people having larger scale deployments, "ask Apple" question. 3 or 4 groups have studied this. - BT: invites ALTO people to attend the PANRG session - RY: do you assume path fixed or that you get properties helping to compute it? BT: is another control loop, we don't assume path is fixed, good question, we need to have control loops (network traffic engineering and endpoints trying to see where to send its traffic) talking to each other and convergence; this is an engineering pb we need address. - Raphael univ campenas.: there is also a network slicing group, are u going talk to them on this? BT: I don't know yet what they do, net slicing is a very large topic, they have a smaller focus, but I think we will. ---------- WG draft Unified Properties for the ALTO Protocol: presented by Dawn Chen https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-unified-properties-00 Discussion - chair JS: like approach (of defining new domains in other drafts), we want to move this work forward. - DC on ALTO Entity Domain (AED) registry: ID is the same (than ALTO Address Type (AAT) registry), address encoding is the same and AED registry allows prefixes. Consistency can be set automatically or manually with specified relationships. - JS: question on auto consistency: if new domain that does not match with IP address, you can create dummy domain with whatever address or the other way round? RY: In the current design we want everything with an address to be in a domain. The problem is in the other way round, for example a PID or ANE has no traditional address type. JS: so what do the authors prefer? comments in the room? RY: we try to go for mutual consistency and look at guidance from IETF WGs or IESG? What's the process? Question is automatic or manual? JS: I have no answer for the moment. How do we move fwd? Choose manual and see what the IESG feedback is? RY: good idea JS: no strong opinion against in room. To move fwd why not do that? Sebastian Kiesel on meetecho: do you mean update RFC 7285? RY: we no change base protocol. It only has AAT registry. AED registry doesn't touch base protocol. DC: we only address relations between AAT and AED registries. Vijay on meetecho: we probably need speak to AD. 2 issues: what makes sense technically? what is the AD and IANA view? JS: we need to see feedback ahead of time and clarify Sebastian: in other words if I don't care about the UP draft and want register new address type in 7285 do I need to care? JS and DC: you don't have to care Kai Gao: you care about UP when you want yo use property pmaps. JS: good progress, so we have still 1 outstanding issue on which we need clarify in following weeks with IANA, AD, and Vijay and list discussions. JS: this document is otherwise well polished ---------- WG draft ALTO extension Path Vector: presented by Dawn Chen https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-path-vector-00 Many edits and clarification on use case. New ANE domain moved from Unified Properties. Main issue: coupling PV cost map and properties of elements returned in the PV response. Different from relations between cost map and network map. We need directions (on PV Cost Map and ANE property map (in)dependency). Discussion - RY comment: we have a solution and look for decisions - JS: you prefer one? - RY: we look for WG expertise and will send e-mail around. Currently we go along for independance, where objects can live by themselves and are composed in "multipart". In YANG there is an option for this but this breaks the protocol. RY: also we removed ANE from Unified Properties because no dependency between both. - JS: as chair I support that. We need solve this issue. We are pretty close to WGLC besides this. ---------- WG draft ALTO incremental updates using SSE: presented by Richard Yang https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-incremental-updates-using-sse-02 Design stable. Lots of re-wording revisions. 1 Use Case across whole doc. For example "update stream" concept, "update message" conveying data update full vs piecewise change. Guidance on data update choices. - JS: like the word "stream", is clearer. RY: we need WG feedback on server response to Client requesting "removal". Discussion - Jon P: we sure need ask an http expert. - JS: very close to WGLC ---------- WG draft CDNI footprint and capabilities advertisement using ALTO: Shawn Lin remotely https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-cdni-request-routing-00 CDNI FCI ALTO service is now CDNI FCI Map. Added IANA considerations: Footprint type registry, new AED registry entries: ASN and Country Code, FCI property registry. - JS as co-author: we introduce new map and need update error handling vs base protocol. - SL: we look for WG feedback and would like to go to WGLC Discussion: - JS: as co-author, I see no major issues from WG. Not that close from WGLC though. ---------- ALTO Cell addresses: presented by Sabine Randriamasy https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-cellular-address-in-alto-00 Proposed as a separate draft but is one case raising discussion on relations between AED and AAT registries in the UP draft. Presentation focuses on that. Cell addreses aloow expanding beyond IP towards 5G. Discussion: - RY: there is a web site where you can look up where AED and AAT are registered. Is that the question? - SR: issue is that WG docs specify how you define information services and use it. But we miss something indicating to people the existence of a given type. For ex. when reading UP, how can I find out there exists something like ANE? - RY: maybe some kind of reverse index ---------- ALTO supported multi-domain orchestration : presented by Danny Lachos Perez https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-based-broker-assisted-multi-domain-orchestration-02 "ALTO-based Broker-assisted Multi-domain Orchestration - 00" This is new work. 5G scenarios need multi-domain orchestrators. Property map and filtered cost map extension. Discussion: - RY : cool. We have related work, CPU prop. For CPUs they (large operators) get confused if different providers have different prop values, we will discuss this offline - Xian meetecho: given Src-Dest pairs, why need to discover all possible paths between them? is it to ctrl the path? Danny: you need to retruen all possible AS path between S and D. ---------- Unicorn updates: presented by Jensen Zhang https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-unicorn-resource-orchestration-for-data-analytics-00 "ALTO Use Case: Resource Orchestration for Multi-Domain, Geo-Distributed Data Analytics" Jensen presents updates on this work. There was a demo at SuperComputing'17 in November. Added extended support to interdomain privacy. Summary: would like to add this document as a use case to the WG charter Discussion -JS: you obviously put a lot of effort in this doc. it is not a WG charter right now, we need see how to move this fwd. ---------- Flow-based Cost Query: presented by Jensen https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-alto-flow-based-cost-query-00 Removed advanced flow-based query. Issues: how to achieve unified query model? how to resolve flow attribute conflicts? Next steps: get WG feedback, move to WG item Discussion - JS: WG item not at this time, we need finish with the milestones ans see what happens with the WG after. I see there is lot of interest in new work. ---------- Routing State Abstraction: presented by Kai Gao https://datatracker.ietf.org/meeting/101/materials/slides-101-alto-draft-gao-alto-routing-state-abstraction-00 Algos for ALTO path vector compression. Concrete implem of PV extension, useful for compression and privacy. Summary: target = informational track, next steps: adopt as WG doc, get WG feedback Discussion - JS: as chair I see usefulness for compression of PV. We need to see if we re-charter the group or not. Agree with Mirja to finish with the WG items. If we re-charter this is a possible WG item. ---------- Chair wrap-up JS: I'm happy we did all the presentations, giving good view of the WG status. Vijay is looking fwd to meeting in Montreal.