Minutes for DMM at IETF-92
minutes-92-dmm-1
Meeting Minutes | Distributed Mobility Management (dmm) WG | |
---|---|---|
Date and time | 2015-03-26 18:00 | |
Title | Minutes for DMM at IETF-92 | |
State | Active | |
Other versions | plain text | |
Last updated | 2015-03-30 |
minutes-92-dmm-1
DMM Session Notes Tuesday: working team1: Exposing mobility state, Alper Mobility Exposure and Selection Presentation summary: DMM API between application and mobile node: Extensions to RFC 5014 – 3 distinct flags added (fixed, sustained, nomadic) yegin-dmm-ondemand-mobility - 3 new flags, to convey between app and IP stack. Liu-dmm-mobility-api - Defines 2 flags, could not understand semantics on call. But it is missing 1 flag and dyn config of IP address Request to adopt this as WG draft. Comments: Dapeng: fundamental question is what is the relation of this to dmm? Does it solve the problem in dmm charter? Alper: going back to a year ago; on demand use of mobility. Dapeng: some extensions proposed, but how is it used? Alper: draft describes use – application and stack behavior. Dapeng: don’t mention what is the problem solved in dmm scope. Alper: this was gone over in previous IETF. Dapeng: there is a gap between charter and solution Alper: what is it? Dapeng: 3 flags added – but there is not related to dmm Sri: this is an enabler for avoiding host routes, tunneling, it is for smaller cells. The idea is to enable the client to enable obtaining new IP address – multiple IP addr, and client intelligently switches (awareness). Dapeng: agree with the statements, but how does the flags help? Alper: flags help applications indicate intention to stack Alex: how does this happen in network Alper: another draft indicates how it happens in network. Danny: will be presented on Thursday. Jouni: this is one enabler – application to stack part. This is independent of other network signaling – IKE, DHCP etc. CoA, HoA in MIP etc. In DMM some addresses are OK and need to know what kind of addr is being requested. Danny: is it now clear? Dapeng: not fully Danny: this is a part solution. Some apps need sustained addr, and others don’t. This draft identifies different services and lets the application communicate to stack. Other draft explains the stack to network part. Dapeng: agree in general, but not specifics. We need to also solve dynamic anchor problem Danny: it is not expected to solve dynamic anchor problem. Alex: Two comments: One - IP addr autoconfig – don’t see a parallel elsewhere in IETF Alper: RFC 5014 already defines 2 options Alex: doesn’t this satisfy the requirements? Jouni: prefix coloring work needed Alper: prefix coloring comes as network-part solution Alex: skeptical. If dhcp wg gives type for addr, then this API makes sense. They should progress together – can WG adoption be made in that case? Jouni: more presentations on Thur. Jouni: during charter, there was agreement for the need of such protocol as prefix color, etc. This is part of solution. Alper: all pieces present is not needed for WG adoption. Alex: 5014 options not implemented in Linux. Alex (comment 2) – do apps run on mobile node? Alper: could be on router. Alex: cannot be on router. This needs to be stated in draft. Sri: we do run applications on mobile routers. Marco: apps which send request and expect an IP address config in response – how can application decide? ' Alper: application can decide – e.g., chat, etc. can indicate to stack. Xinpeng: will the app developer be confused - because they do not care about network address – ie. Will they use the options? Alper: apps that want efficient networking will adopt. Xinpeng: is the prefix defined – should it be defined here in dmm, or.. Alper: considered here first, and then taken to different normative WG (dhc, 6man). Jouni: the working teams are open and can be joined, reviewed by anyone. Alex: agree that it is open --- on the other hand, it is in IETF meetings that decisions are made and should not hold that against one who did not participate. working team2: Enhanced anchoring, Anthony/25min Presentation summary: Enhanced Mobility Anchors work team report Enhanced anchors for moving networks AAA arch for DMM Moving IP address Allocation of IP address to mobile mode – HoA, HNP etc Functions of an anchor Anchor selection and switching Anchoring IP address Comments: Sri: is prefix delegation initiated by client? Anthony: not by the client. Its from one router to another. Sri: what protocol is used? Anthony: prefix delegation protocol, or SDN, control plane. Anchoring session (with and without IP address change) Anchoring of the session in addition to the IP address implies move of session anchor Sri: what is a session anchor? Are you routing an IP flow through one path, and session at another point? Anthony: you try to make the session go through the anchor. Sri: how is the forwarding plane able to steer the IP flow through one path, and others to another path? Some kind of stitching? Anthony: session anchor moves independent of IP. Marco: Mobile node has 2 IP sessions, one of which is on AR2. Is IP1 deprecated? Anthony: (referring to slide 11) - anchoring only the IP address on AR (right side). Sri: not following this. Session re-anchoring – how we move the subscriber session to another node is not related. Anthony: we do not have a definition of mobility anchor for session, and that part is not enough. Sri: what is a session A: it is a flow Sri: seems like a mixup of terms. Kent: clarify session – seems to be that IP session can move independent of IP address Anthony: [walk through slides again] Kent: transport of flow to that anchor, topology, etc., needs to be clarified Kent: session is not at the flow level – can do tunneling. Cannot route at flow level. Sri: route optimization is different. CoA and HoA are present in the headers and that’s how it is handled. Anthony: this is just at high level, it is just an analogy. Trying to harmonize the different proposals Sri: IP address and sessions are on different nodes – that’s not clear. Jouni: there is a problem with terminology because people cannot flow – need clarity on the terms. Describe with existing technology so that it is not difficult to follow. Show what the mechanisms are to move between the nodes. Dapeng: is the draft a consensus of the team? A: there are comments from team, but open to listen to more comments. Thursday: working team3: Forwarding path and signaling Management, Macro Presentation Summary: Functional decomposition & FPSM is the scope. Protocol and mechanisms that work across stds. FPC client and FPC agent use common protocol semantics for policy config, queries and notifications. Generic policies and forwarding rules are defined – logical port, grouping concepts. Next step is to adopt information/ data model. Comments: Fred: wrt PMIPv6 as example – could AERO be another one of them? Marco: it is orthogonal. AERO protocol can use it too. Carlos: in favor of adopting Georgios: in favor of adopting John: what is the relationship to Forces, OF? Marco: focus here is on mobility and semantics needed. Next step would be to apply to concrete implementation like OF. Xinpeng: why is QoS included in the attributes? Isn’t this about mobility management. Marco: there are extensions for QOS and mobility management as in PMIPv6 QoS. Jouni (Chair): is this mature enough to be adopted as WG doc? In Favor? [many] Should wait and see for more content. [1] working team4: Distributed mobility management deployment models and scenarios, Sri Presentation summary: CP-DP separation. Aggregation of control plane can provide elastic scaling. Selection of GW based on session requirements (e.g., DPI for some sessions) Mapping of DMM functions including aspects of splitting functions into CP/DP, mapping to different architectures. It provides the ability to pick a data plane anchor (DPA) on demand. Security model here includes – IPSec, IKEv2. Deployment models explained. Comments: Danny: CP is centralized, DP distributed. One control entity and many DP entities? Sri: if the question is - do we decompose the control plane, then consider that we could have VMs, etc. Danny: should it be part of this doc? Sri: this can be part of informational text. Dapeng: is there session continuity support for anchor relocation? Sri: yes. Hui: can you do chaining for DPA? Sri: could have root anchor doing one function, and then [interrupted ...] Hui: IP acceleratio, DPI, etc. on data path is the question Sri: I see - service function chaining – not captured yet.. Hui: these functions are inside a GW for example Sri: yes. Gi-LAN or similar functions in the data plane needs to be told where to steer the traffic. Anthony: why distinguish between node and anchor? (and not have anchor-anchor) Sri: in today’s network, session is anchored at LMA at home, while not at MAG. Anthony: so its where IP is anchored. Sri: right. Marco: data plane functions may be chained, and this needs to be considered. Sri: yes. draft-matsushima-stateless-uplane-vepc-04 Satoru Presentation summary: changes in the new revision - Forwarding Policy Configuration Protocol as a means to expose Control-plane states from mobility management to BGP speaker. Next steps – outline how this is applied to work item 3, 4. Comments: Dapeng: is this aligned with arch as Sri presented? Satoru: not sure, but this work was started before and trying to align. Sri: it is aligned, but we should ask for more comments Fred: the presentation referred to BGP work from Boeing. Will you refer to RFC 6179 for DMM also? Satoru: it would be different purpose vs. the use of BGP here (as vertical protocol). Marco: this fits into work item#4. It is very concrete implementation. Jouni: not ready to ask the question of adoption call today, but would like to see more people work on this. Jouni: who is ready to work on this draft? [many] Jouni: who thinks this does not fit? [no one] Alex: is this implemented in some platform for research/students, or is it in a proprietary platform? Satoru: different parts in the solution – EPC, BGP, etc. The extension of BGP is available, but EPC is not. Alex: EPC is not something implemented at universities. draft-moses-dmm-dhcp-ondemand-mobility-00 Danny Presentation summary: address types – how to extend IPv6 client to convey to the application what the client is requesting IA address option carrying source IP. New option: IP continuity service option to carry the type of the IP address. This work will need to be taken to DHC group after agreement in dmm. Comments: Fred: as a mechanism, this is fine. Fixed addr, nomadic addresses are returned – how is it assigned to the right interface? Danny: stack needs to keep track. Fred: how does the network know which address is nomadic/static etc. Danny: network does not care. When the client moves to new location, nomadic addresses cease to exist. Network does not nave to support moving it, tunneling etc. Fred: this is a simple extension, but there is a lot of detail underneath that is not shown. Sri: a comment on how apps make distinction of address – there will be an impact to RFC3484, right? Marco: puts assumptions on DHCP and that puts some impact on anchor selection Danny: yes. [Draft Yegin adoption call] Jouni (Chair): is this a good basis for first draft for the application – stack. In favor: [many] Oppose: 3 Jouni: rough consensus… Brian (AD): who is willing to put in work on the doc? [Response is about the same group as those that support] Chair (Dapeng): please make clarifications to document based on questions on the mailing list. draft-sijeon-dmm-use-cases-api-source-00.txt Alper Summary: Further extensions to the dmm api extns draft Flag for new application to indicate preference for new address Comments: Danny: which application would prefer to be anchored to a far away anchor? Alper: [interrupted ..] Danny: it asks the stack to ask for new addresses – how does the stack know, and how will this not be abused? Fred: if the sustained address is always close, then would this be needed? Alper: these need to be considered Jouni: can this be added into draft-yegin? Alper: yes. draft-yan-dmm-hnprenum-00.txt Zhiwei HNP renumbering based on RFC 7077 Comments: Jouni: is this part of maintenance part of dmm (PMIP related)? Presenter: yes Sri: technically makes sense. RFC 7077 makes sense.. Alex: will the renumbering imply breaking session continuity? Presenter: session would be restarted Jouni (Chair): please comment on the mailing list in support or against, and bring up the issues. draft-templin-aerolink Fred presentation – no comments. Discussions: Charlie: with regard to mobile node id draft – the intention is to request for adoption. Jouni: this is related to the maintenance part of this group. Alex: agree with most points, so an adoption would make sense if it is in charter. Alex: Is this open for suggesting new identifiers? Chair: yes Alex: developing new id for car. Chair: bring it in to the working group. {next discussion} Sri: Pierrick has a document on hybrid access that needs follow up. Chair: don’t recall any discussion on thread.