IP over IEEE 802.16 Networks (16ng) WG -------------------------------------- ========================================================= Special thanks to Dave Thaler and Alex Petrescu for their effort on the notes ========================================================= Agenda bashing - chairs 16ng Problem Statement - Junghoon Jee IPv6 transmission over IPv6 CS - Jinhyeock Choi IP transmission over Ethernet CS - Max Riegel IP Deployment over WiBro - Sang-Eon Kim IPv6 over WiBro Report - Eun Kyoung Paik 802.16 fixed/nomadic IP deployment - Max Riegel DAD Optimization Using Enhanced ND - Frank Xia o Chairs - WG report WG approved june 29, tech advisor is Max Riegel who's active in wimax forum Liaisons from 802.16 and wimax are under consideration Interim meeting in September 11-12 or 12-13 Logistics: Seattle, Seoul, Germany - tbd Polled for venue and dates, chairs didn't announce counts o Max Riegel - 16ng relationship with related efforts Wimax is subset of 802.16 No new features can be added Two worlds: fixed / mobile 16d OFDM-256: IP*oETH CS / n/a 16e OFDMA: IP*oETH CS / IP*CS o Junghoon Jee - problem statement Subscriber station (SS) = Host #1: NBMA link no native bidir mcast support No direct communication between hosts, always between host and BS 48-bit mac addrs not used for transmission #2: Ethernet cs doesn't provide Ethernet-like functionality #3: Air resource consumption from the previous mcast/bcast based ND/ARP #4: Power consumption caused by waking up of dormant subscriber station #5: No single mandatory CS on the subscriber station #6: MLS issues when using IP CS for Ethernet-like IP subnet model Network models: IP subnet over 802.16 P2MP links Ethernet-like: hosts have common subnet, both CS's applicable IP CS: need to resolve MLS issue Ethernet CS: BS does bridging 3G-like: SS's don't share prefix IP CS seems more applicable Q(Myungki): why do we want to do something to negotiate CS? A(Daniel): we don't need anything new in this WG Q(Dave Thaler): so you really mean (putting this together with Max's presentation) that fixed 802.16 should use ethernet like subnet model, and mobile 802.16 could use either, right? A(Junghoon Jee): yes Pat Calhoun: WiMax forum is not an internationally recognized SDO, it's a marketing organization Q(Myungki):need to discuss about subnet and deployment models. There can be the following IP network models over 802.16 - 1) single/mobile device deployment model - it is WiMAX/WiBro 2) bridging to general ethernet model 3) 3g-like {RFC 3316) deployment model A (Junghoon Jee) Yes, if necessary, we will consider them, too. Chairs asking for adopting problem statement doc as a WG document, no objections in room. Final confirmation on the mailing list. o Jinhyeock Choi - IPv6 over IPv6CS Shared subnet Offlink model in Ras, w no redirects No TTL decrement No multicast support "WiMax specific solution" to run IPv6, could be developed into generic solution No unique LL addr All traffic via AR Hosts on-link nbr is only its AR Connection between host and AR (not BS) All hosts on same link (Ethernet-like model) Prefixes advertised without on-link flag (offlink model), no redirects No mac in RS so initial RS from unspec'ed address must be replied to per AR-port Address resolution could be skipped since no on-link routes NUD is just between host and rtr Router responds to NS out of its own cache, which is populated from DAD Q (Dave Thaler): cache population via DAD isn't sufficient, it doesn't work for several reasons: pkt loss, anycast, Dadattemtps=0, bad "opt" of LL but not global, etc Q (DT): what about multicast A: not supported Q (Gorry Fairhurst): what about UDLR? A: (Myungki) the current doc is stable and has a good shape for WiMAX, but may be not for general IPv6 CS solution. IETF document should have broader view rather than this. WiMAX is one of the current deployment examples. So, 16ng document should cover other deployment models, too. o Max Riegel - IP over Ethernet CS (Fixed) Issues with Ethernet on Wireless: Multicast must be transmitted multiple times Limited terminal power, terminal has to wake up to receive packets Per-NL protocol classification causes different CSs Must learn about MAC/IP addresses of the MSs (hosts) to prevent unwanted transmissions Q (DT): so if it's not in table, do you flood or what? A: yes, cache is just an optimization Standard learning bridge - 802.1D But closely follows spec for IPv6 CS o Sang-Eon Kim - IP Deployment over WiBro WiBro = wireless broadband Wibro is subset of mobile wimax (max's slide #3) ??? slide appears to show Multiple CS's for IPv6 - bearer, control, and management WiBro is IP*CS (not Ethernet) Q (John Loughney): are you defining new RTO algorithms? A: no What's difference between CDMA and WiBro? A: o Max Riegel - 802.16e Fixed/Nomadic (Lower-left quadrant) o Eun Kyoung Paik - IPv6 over WiBro IPv6 CS only TTA std in korea, commercial svc began June 29, 2006, but definition still in progress Still todo is multicast and mobility o Frank Xia - DAD optimization DAD NS, addr cache lookup in BS/AR, forward to duplicate MS (only), let MS reply with NA If not found, discard? When AR finds that there is no such IP addrs in its "authoritative" cache, RA is send to indicate uniqueness of the address, and on receiving the positive reply, the MS will stop waiting and use the IP address immediately Dave Thaler: not needed on Ethernet CS, can just use Optimistic DAD. Dave Thaler: On IP*CS, the assumption that the router knows all on-link addresses based on DAD does not hold in the IPv6 architecture/protocol. There are at least 4 reasons, probably more, why this is a faulty assumption. o Chairs: closed