mobopts research group IETF 83 Monday, March 26, 2012 1300-1500 Afternoon Session I ============================= Chairs: Suresh Krishnan Rajeev Koodli Suresh: - Purpose is to check if there is any interest in IP mobility and have a discussion. - Agenda is introduced. 1) Jari presents Analysis of problems to get IETF mobility protocols to a deployable state - Summary of what has been defined already - Problems and reasons why Mobile IP is not being used for its purpose today. - Engineers' view is different from users' view - Question: Who you consider as the user? - Jari: That is part of the problem. But here is the user from the point of view of who deploys, maybe the operator. - Engineers talk about bits and users talk about value/cost - MIP competes with other layers mobility solutions - Applications have changed to survive address changes, because they didnt have MIP, due to NATs so they were forced, only few apps need stable IP address - Unclear motivation - Implementation affects many parts of the stack - Jari: This is personal opinion about the issues I see for the deployment. It would be more correct to say IP mobility is not being deployed the way that IETF defined it. - Comment: IETF tries to be agnostic regarless of the access technology - Jari: Agree, but competition from the other layers shrinks the space of work for us. Efforts devoted to security, optimization, etc, but those are not the problems why the protocol is not deployed. The question is do you actually need IP mobility? - Comment: Mobile ip is great but the only app that needs it is the one that does not support some perturbation in packet delivery, but most apps dont need it. And the app that needs it already has the solution implemented itself. - Charlie: People like to say that voice service is already so bad that the improvement from MIP is irrelevant. - Suresh: Systems that use GTP are integrated with all the other components, like billing, location, etc, so it is hard to implement MIP. - Julien: We recreated the same with PMIP, and it worked great. - Question: Are you in the opinion we dont need to think too much about IP mobility? - Jari: I think the need is there, but can't we supersede the other solutions that are alredy there - Comment: but you suggested is better to do it at the app layer - Jari: If we were to think everything again it will be a combination of MIP with other things, not necessarily at the app layer - Jari: There is a clear problem that is not being addressed by the other layers. The question is, is there enough market, enough users. - Lars: Stable IP is nice, but even with that the app still hiccups when handover occurs. If you look at packet loss metric is successful but from app layer is different. 2) Basavaraj presents Rethinking IP Mobility - Mobility does provide a significant value proposition, but the primary need is Internet connectivity, then IP mobility is a value add. - Question: What is the reason for not having all the mobility features in the IPv6 stack - Charlie: IPv6 itself has not been deployed, and MIPv6 was designed to go along with IPv6. Cellular systems already had their own solution, so they were not that much interested on this. - Jari: The spec for mobility came out later than the spec for IPv6. And it is a different function so it makes sense to have them separated, as this is an extra feature. - Suresh: We put it as a requirement, and implementers should have it, but it didn't happen. Nobody really specified it as mandatory. - Raj: In order to implement IPv6 you dont need to have IP mobility. - Devices are having more interfaces, so how do you handle the handoff between different interfaces - Julien: Handoff between different techonologies has also been solved, there are 4 different protocols. You can use GTP when roaming to WiFi. - Raj: I know there are solutions but there are things that still need to be done. You may provide a stable address but apps do not necesarily see continuity. - Look at the problem from the top-down perspective. There are few apps that really need IP mobility. Most apps are short-lived. - Idea is to present to the apps another interface, so if the app needs a stable address it has the option to choose that interface. - Jari: Apps require code changes? - Raj: Apps will determine what interface they need. You just provide this interface and if the app needs it, then it is available for use. - Lars: If the only benefit is a stable IP, then is not enough. The general mobility problem has been solved already. - Suresh: If the app wants mobility, it will ask for a stable IP address (like a home address) and then you will be bounded to the stable interface. - Question: How the app decides which interface is stable? - Sri: The information about the prefix may include the characteristics. It is prefix coloring. - Question: This is within the host stack, then why do we need standard? - Suresh: Because the apps developers need to work with a standard interface so that it is not different in every platform. - Juan Carlos: We started talking about mobility, but this is more about providing different services to the applications. We have a similar discussion in the MIF group and at the time we discussed if that is something that should be discussed in MIP. - Comment: What you are doing is not a mobility solution. It is just to expose more information to the application. The other thing is how much do we want to expose to the application. - Raj: There is no question that there are many solutions already for providing mobility. We are just saying that some information should be exposed to the app. - Question: Why would we want to standardize this API. - Lars: There is an assumption here that the prefix coloring will give some information to the app, but the reality is that you don't know anything until you have the app running. A normal app will work with MP-TCP, but for media you may need something else. For general apps is difficult, but for media where you know codecs and other things you may need it. - Raj: There is a space of apps that will benefit from this. - Suresh: The take-away from this is not to provide mobility to everyone but instead to provide more information so the app can make an informed decision about what interface to use. - Question: How does the operator can control which interface is the user going to use? - Suresh: There is work to put some information in DHCP. The policy table is going to change. Add a multiple site policy. Then the host has to put all this together before making the decision. - Question: How does the operator control that? - Suresh: The app in the device decides. Windows already does it, and many others do it too. - Charlie: Why to discuss IP mobility if there are solutions in other layers. If you want to revisit this why don't we bring GTP as well. I made some presentations before in MEXT about why I think is a bad idea that Mobile IP does not support GTP. It was very difficult for cellular system to evolve to Mobile IP. - Julien: GTP complexity cant be compared to MIP. GTP is not a mobility protocol. - Suresh: If MIP would have to do all that GTP does, it will be even more complex, like location, charging, etc. I believe the problem is in the control plane. - Charlie: We also looked at the split control/data plane problem. You are right, GTP is much more than a mobility protocol. - Julien: 3GGP changed the GTP signalling to make it compatible with LTE, and MIP could have happened at that moment, but it didn't. - Comment: They did it just because MIP couldn't do all the things that GTP was already doing. Peter presents Mobility Management in a Flat Architecture - Drawbacks of Existing Hierarchical Tunnel Solutions - Elements of a Flat Wireless Internet Service Provider - Suresh: So would you have to re-authenticate every time you move? - Peter: Yes, and we also propose a fast authentication mechanism. - Suresh: In previous experiments it was shown that worst case scenario is that you need one entry per MN - Peter: You have to subdivide according to regions - Jari: Thinking about the scale. Today's routers hold between 1 to 10 million entries. If that is the number of mobile nodes you can have then you have a limit of say 1 million devices. What if you roam from a set of BS to another one? - Peter: Then you have to open a MIP tunnel. - Jari: The update frequency times is something that needs to be defined. Those are difficult-to-achieve requirements. - Kent: This was dicussed before in 1988 - Peter: Yes, but we have now more experience building these networks. - Question: Are you looking to LISP as well? - Suresh: In the previous MOBOPTS meeting we had a presentation about how LISP does mobility. So you can have a look to that. - Albert: The real routing updates will be less than that because not all the nodes need mobility. - Marko: Is the idea of this to use BGP to update the MNs state even though they are not active? - Peter: No, you only use BGP when you need to punch a whole in the routing table because you moved to a new network. 3) Dapeng presents DMM overview -Brief explanation about DMM drafts 4) Discussion - Suresh: Is there any point on looking at IP mobility anymore? Is there any research issues left? - Lars: If you look at the charters they are completely out of date of whats going on. On top of that, in the mailing list and meetings there hasn't been any energy going on, the meeting today is an exception. What should a research group in this space focus on? the publications in the area have been declining. - Albert: We should keep discussing what happened to MIP and why happened. That discussion should take place and then decide, whether it happens here or at the IETF. - Raj: I disagree with Lars, I have seen people bringing research and presenting in this group, and that has been useful. Having people doing simulations and prototypes about the implications, etc is still useful. - Suresh: Who is doing research in IP mobility - Lars: Then bring it and present it. I have looked at the mailing list and there have been less than 20 emails in the last year. - Suresh: Anyone that has a proposal please send it to the mailing list and we can decide. Lars will decide if the WG continues or not. I welcome you to bring over any problem you are researching.