gomm 10july2006 Eric Njdedjou EN EN presents network global mobility management - a problem statement draft-njedjou-globalmm-ps-00.txt How many people looked at the draft - a few hands up. a slideset is presented. Vijay Devarapalli: VD VD: minor comment, previous slide, from user persepctive: good mobility experience. EN: absolutely, I'll have to add it. Gerardo Giaretta GG GG: two main problems that are different. Operator point of view I agree . There are two requirements: 1st - keep control by the network, 2nd - keep minimal host support, minimal radio link. 1st part can be achieved with a host-based protocol e.g. mip, 802.21 with them you can keep control in the network EN: from a functionnal point of view, yes you can. But at the cost of adding several layers of mobility control (MIP, 802.21), thus not easing the task of designing scalable architectures and stable mobility systems EN: network-initiation control scheme. Junghoon: J Junghoon: vertical handovers? Do you use "global mobility" to present vertical handovers? Or netlmm? EN: no, not netlmm. The global mobility is effectively often described with vertical handovers J: global mobilmoity doesn't necessarily mean vertical handover. EN: ABsolutely. Everything depend on the topology that is behind. global or local depends of how far in the routing topology you have to go boack to change routing. James Kempf: JK JK: existing protocols in this slide? EN: no JK: which protocol is in blue? RFC? 3gpp? EN: The protocol in blue could be the 802.21 MIH protocol EN: after the step of exchanging information for handover control, you have to engage the se session continuity protocol . JK: red lines is MIP right? EN: yes. JK: architecture, IETF doesn't do . IETF rather oes specific parts. If the intent is to design big protocols that have all the implications as shown, then IETF is not the right place. EN: The goal is to have IETF designed the specific interfaces it is pertinent for. These are the interfaces for IP mobility. However, IETF has to think which implications these interfaces have when considering the global architecture GG: I tempt to agree with JK, if you put everything in consideration, how to do the IP mobility, how to provide hint to mobile, that means you're designing a system, not a protocol. EN: what if IETF design the IP layer protocol with everything else in mind? That does not necessarily entails designing the whole system EN: to initiate operations from the network for examle, you have to have a specific order of operation for your mobility protocol GG: I guess you really need to have an archi in mind, we don't do system design, just protocols. Jari Arkko: JA JA: statistics or performance analysis would be needed to illustrate the problem- why do we need such architectures?. If you consider any model... is there any analysis on signalling you have? EN: wide scale analysis is not possible because there is no deployment. Only lab trial is achievable. Even without an assessment of the perfromance , you can still anticipate that because of frequent movements between heterogeneous systems, that you may need to low down the amount of sugnalling or that kind of stuff JA: often different types of links. EN: The sigalling issue is not be the corner point. The corner is the complexity of the whole system to adddress mobility Stefano Faccin: SF SF: IETF does not design combined pieces. Confused also about what is global mobility, when you talk about vertical handover. mobility between heterogeneous domains doesn't happen very often. EN: yes but vertical handovers will not necessarily happen between domains. Thise can happen within a single domain. Agree that talking of global mobility might be rather confusing howver the problem of changing between systems still remains. SF: difficult to discuss a problem about global mobility when we don't know what it is Vidya Narayanan: VN VN: mobile node doesn't have a HoA and/or CoA? Are you envisioning network-based local mobility management ?- how is it feasible? Two interfaces active simultaneously? Mobility of these two would be managed separatedly. Is that multi-homing? EN: I don't see how why it is not possible to have two active interfaces simultaneously and do global IP mobility without MIP. Two interfaces can be active but not used simulatenously. I am not saying all of the concepts of MIP are to be changed to address the problem. There is alsways the need in this problem statement to hide IP address change for the sake of continuity of sessions. VN: each address is treated independently. If endhost runs MIP then... explains MIP and two HoAs. VN: if the network is doing it it has no idea.. EN: transition between two domains VN: wlan and cellular - do you assume same IP address? EN: no VN: confused. EN: what do you mean?... The point is not about having two interfaces with addresses treated independantly, I don't get the point. VN: netlmm model handles IP mobility for a given IP address, if the device has multiple addresses. no way to manage that without a hoa concept. Discussion EN-VN on multiple interfaces is to be carried out off line. Marco Liebsch ML ML: if a subsequent problem statement is presented, design goals, solution should follow. what solution focuses on? keeping policy decision point in networks for handover targets? or retrieve measurements from mobile? which area? EN: mobility management is complex in itself. You have handover management to be realized, routing to be changed from one part to antoher, the point is to see how to avoid burden of being too complex while realizing these functions ML: the problem is related to integrating missing functions to meet requirements, not to define new protocols? EN: end goal was to see if there is necessary to design something different Qiaobing Xie: QX QX: confused on people reactions... IETF should ask itself couple questions about the problem just presented. IETF protocols and work happening o, mobility is very good for conference, publications.... But for global communications systems IETF does not have not enough. IETF is not very good at that. Here's a good perspective of what is the complete mobility problem. Nation-wide or global is another requirement. I don't think this presentation requires IETF to design architectures or systems, but tries to identify the new pieces that should be created. Slide with blue-red vertical lines: this is typical way of doing mobility management in GSM systems Some new protocols over IP are required to do this. This is something very valuable to think about. This is not requiring IETF to define archi, just show the big picture, the point is to find what's missing to be covered by OETF, what can be done in the next 3 years. JK: cdma/gsm doing this kind of thing, there are protocols doing this, e.g. GRX. GRX has a lot of implementation problems, because it was not designed by designers experience. Collaboration between operator sometime help solve the problem. But IETF is much more than just operators cooperating. Relationships between operators and IETFers are more... What you're proposing is another inter-operator interface... not sure that's a good match for IETF. Look at the boeing connexion solution. not scalable of doing global mobility. Only because works because only a few airplanes. EN: GRX not for mobility. GRX is for exchanging agreements. more dynamic mobility procedures I'm talking about. EN: setting up a page. GOMM web page setting up. Problem that was raised was not necessarily fully tight to what IETF is to deliver, I'll have to go in better details. Please provide comments.