J. Kempf, Editor Internet Draft K. Leung Document: draft-ietf-netlmm-nohost-req-01.txt P. Roberts K. Nishida G. Giaretta M. Liebsch Expires: October, 2006 April, 2006 Goals for Network-based Localized Mobility Management (NETLMM) (draft-ietf-netlmm-nohost-req-01.txt) Status of this Memo By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware have been or will be disclosed, and any of which he or she becomes aware will be disclosed, in accordance with Section 6 of BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract In this document, design goals for a network-based localized mobility management protocol are discussed. Table of Contents 1.0 Introduction.....................................................2 2.0 Goals for Localized Mobility Management..........................2 3.0 Security Considerations..........................................8 4.0 Author Information...............................................9 5.0 Informative References..........................................10 6.0 IPR Statements..................................................11 7.0 Disclaimer of Validity..........................................12 8.0 Copyright Notice................................................12 9.0 Appendix: Gap Analysis..........................................12 Kempf, et. al. Expires October, 2006 [Page 1] Internet Draft LMM Goals and Gap Analysis April, 2006 1.0 Introduction In [1], the basic problems that occur when a global mobility protocol is used for managing local mobility are described, and three basic approaches to localized mobility management - the host-based approach that is used by most IETF protocols, the WLAN switch approach, and using a standard routing IGP to distribute host routes - are examined. The conclusion from the problem statement document is that none of the approaches has a complete solution to the problem. While the WLAN switch approach is most convenient for network operators and users because it requires no mobile node support, the proprietary nature limits interoperability and the restriction to a single wireless link type and wired backhaul link type restricts scalablity. The IETF host-based protocols require host software stack changes that may not be compatible with all global mobility protocols, and also require specialized and complex security transactions with the network that may limit deployability. The use of an IGP to distribute host routes has scalability and deployment limitations. This document develops more detailed goals for a network-based localized mobility management protocol. In Section 2.0, we review a list of goals that are desirable in a network-based localized mobility management solution. Section 3.0 briefly outlines security considerations. More discussion of security can be found in the threat analysis document [2]. The architecture of the NETLMM protocol for which the goals in this document have been formulated is described in Section 4 of [1]. 1.1 Terminology Mobility terminology in this draft follows that in RFC 3753 [3] and in [1]. 2.0 Goals for Localized Mobility Management Any localized mobility solution must naturally address the three problems described in [1]. In addition, the side effects of introducing such a solution into the network need to be limited. In this section, we address goals on a localized mobility solution including both solving the basic problems and limiting the side effects. Some basic goals of all IETF protocols are not discussed in detail here, but any solution is expected to satisfy them. These goals are interoperability, scalability, and minimal goal for specialized network equipment. A good discussion of their applicability to IETF protocols can be found in [5]. Out of scope for the initial goals discussion are QoS, multicast, and dormant mode/paging. While these are important functions for mobile nodes, they are not part of the base localized mobility management problem. In addition, mobility between localized mobility management domains is not covered here. It is assumed that this is covered by the global mobility management protocols. Kempf, et. al. Expires October 2006 [Page 2] Internet Draft LMM Goals and Gap Analysis April, 2006 2.1 Handover Performance Improvement (Goal #1) Handover packet loss occurs because there is usually latency between when the wireless link handover starts and when the IP link handover completes. During this time the mobile node is unreachable at its former topological location on the old IP link where correspondents are sending packets and to which the routing system is routing them. Such misrouted packets are dropped. This aspect of handover performance optimization has been the subject of an enormous amount of work, both in other SDOs, to reduce the latency of wireless link handover, and in the IETF and elsewhere, to reduce the latency in IP link handover. Many solutions to this problem have been proposed at the wireless link layer and at the IP layer. One aspect of this goal for localized mobility management is that the processing delay for changing the routing after handover must approach as closely as possible the sum of the delay associated with link layer handover and the delay required for active IP layer movement detection, in order to avoid excessive packet loss. Ideally, if network-side link layer support is available for handling movement detection prior to link handover or as part of the link handover process, the routing update should complete within the time required for wireless link handover. Note that a related problem occurs when traffic packets are not routed through a global mobility anchor such as a Mobile IP home agent. Route optimized Mobile IPv6 [6] and HIP [7] are examples. A loss of connectivity can occur when both sides of the IP conversation are mobile and they both hand over at the same time. The two sides must use a global mobility anchor point, like a home agent or rendezvous server, to re-establish the connection, but there may be substantial packet loss until the problem is discovered. Another aspect of this goal is that the solution must ensure that connectivity is not lost when both ends are mobile and move at the same time. In both cases, the loss of accurate routing causes the connection to experience an interruption which may cause service degradation for real time traffic such as voice 2.2 Reduction in Handover-related Signaling Volume (Goal #2) Considering Mobile IPv6 as the global mobility protocol (other mobility protocols require about the same or somewhat less), if a mobile node is required to reconfigure on every move between IP links, the following set of signaling messages must be done: 1) Movement detection using DNA [8] or possibly a link specific protocol, 2) Any link layer or IP layer AAA signaling, such as 802.1x [9] or PANA [10]. The mobile node may also or instead have to obtain a router certificate using SEND [11], if the certificate is not already cached, 3) Router discovery which may be part of movement detection, 4) If stateless address autoconfiguration is used, address configuration and Duplicate Address Detection (unless optimistic Duplicate Address Detection [12] is used). If stateful address configuration is used, then DHCP is used for address configuration, Kempf, et. al. Expires October 2006 [Page 3] Internet Draft LMM Goals and Gap Analysis April, 2006 5) Binding Update to the home agent, 6) If route optimization is in effect, return routability to establish the binding key, 7) Binding Update to correspondent nodes for route optimization. Note that Steps 1-2 will always be necessary, even for intra-link mobility, and Step 3 will be necessary even if the mobile node's care-of address can remain the same when it moves to a new access router. The result is approximately 10 messages at the IP level before a mobile node can be ensured that it is established on a link and has full IP connectivity. Furthermore, in some cases, the mobile node may need to engage in "heartbeat signaling" to keep the connection with the correspondent or global mobility anchor fresh, for example, return routability in Mobile IPv6 must be done at a maximum every 7 minutes even if the mobile node is standing still. The goal is that handover signaling volume from the mobile node to the network should be no more than what is needed for the mobile node to perform secure IP level movement detection, in cases where no link layer support exists. If link layer support exists for IP level movement detection, the mobile node may not need to perform any additional IP level signaling after handover. 2.3 Location privacy (Goal #3) Although general location privacy issues have been discussed in [14], the location privacy referred to here focuses on the IP layer. In most wireless IP network deployments, different IP subnets are used to cover different geographical areas. It is therefore possible to derive a topological to geographical map, in which particular IPv6 subnet prefixes are mapped to particular geographical locations. The precision of the map depends on the size of the geographic area covered by a particular subnet: if the area is large, then knowing the subnet prefix won't provide much information about the precise geographical location of a mobile node within the subnet. When a mobile node moves geographically, it also moves topologically between subnets. In order to maintain routability, the mobile node must change its local IP address when it moves between subnets. A correspondent node or eavesdropper can use the topological to geographical map to deduce in real time where a mobile node - and therefore its user - is located. Depending on how precisely the geographical location can be deduced, this information could be used to compromise the privacy or even cause harm to the user. The geographical location information should not be revealed to nor be deducible by the correspondent node or an eavesdropper without the authorization of the mobile node's owner. The threats to location privacy come in a variety of forms. Perhaps least likely is a man in the middle attack in which traffic between a correspondent and the mobile node is intercepted and the mobile node's location is deduced from that, since man in the middle attacks in the Internet tend to be fairly rare. More likely are attacks in which the Kempf, et. al. Expires October 2006 [Page 4] Internet Draft LMM Goals and Gap Analysis April, 2006 correspondent is the attacker or the correspondent or even the mobile node itself is relaying information on the local address change to an attacker. The owner of the correspondent or mobile node might not even be aware of the problem if an attacker has installed spyware or some other kind of exploit and the malware is relaying the change in local address to the attacker. Host-based localized mobility management solutions in which the correspondent only sees a regional address but the host still maintains a local address are unsatisfactory because they still have a potential for malware on the mobile node itself to reveal a change in the local address. Note that the location privacy referred to here is different from the location privacy discussed in [16][17][18]. The location privacy discussed in these drafts primarily concerns modifications to the Mobile IPv6 protocol to eliminate places where an eavesdropper could track the mobile node's movement by correlating home address and care of address. In order to reduce the risk from location privacy compromises as a result of IP address changes, the goal for NETLMM is to remove the need to change IP address as mobile node moves across IP links. Keeping the IP address fixed removes any possibility for the correspondent node to deduce the precise geographic location of the mobile node without the user's authorization, as well as any possibility that malware on the mobile node could inadvertently reveal the mobile node's location to an attacker. Note that keeping the address constant doesn't completely remove the possibility of deducing the geographical location, since a local address still is required. However, it does allow the network to be deployed such that the mapping between the geographic and topological location is considerably less precise. If the mapping is not precise, an attacker can only deduce in real time that the mobile node is somewhere in a large geographic area, like, for example, a metropolitan region or even a small country, reducing the possibility that the attacker can cause harm to the user. 2.4 Efficient Use of Wireless Resources (Goal #4) Advances in wireless PHY and MAC technology continue to increase the bandwidth available from limited wireless spectrum, but even with technology increases, wireless spectrum remains a limited resource. Unlike wired network links, wireless links are constrained in the number of bits/Hertz by their coding technology and use of physical spectrum, which is fixed by the PHY. It is not possible to lay an extra cable if the link becomes increasingly congested as is the case with wired links. Some existing localized mobility management solutions increase packet size over the wireless link by adding tunneling or other per packet overhead. While header compression technology can remove header overhead, header compression does not come without cost. Requiring header compression on the wireless access points increases the cost and complexity of the access points, and increases the amount of processing required for traffic across the wireless link. Since the access points tend to be a critical bottleneck in wireless access networks for real time traffic (especially on the downlink), reducing the amount of per-packet processing is important. While header compression probably cannot be completely eliminated, especially for Kempf, et. al. Expires October 2006 [Page 5] Internet Draft LMM Goals and Gap Analysis April, 2006 real time media traffic, simplifying compression to reduce processing cost is an important goal. The goal is that the localized mobility management protocol should not introduce any new signaling or increase existing signaling over the air. 2.5 Reduction of Signaling Overhead in the Network (Goal #5) While bandwidth and router processing resources are typically not as constrained in the wired network, wired networks tend to have higher bandwidth and router processing constraints than the backbone. These constraints are a function of the cost of laying fiber or wiring to the wireless access points in a widely dispersed geographic area. Therefore, any solutions for localized mobility management should minimize signaling within the wired network as well. 2.6 No Extra Security Between Mobile Node and Network (Goal #6) Localized mobility management protocols that have signaling between the mobile node and network require a security association between the mobile node and the network entity that is the target of the signaling. Establishing a security association specifically for localized mobility service in a roaming situation may prove difficult, because provisioning a mobile node with security credentials for authenticating and authorizing localized mobility service in each roaming partner's network may be unrealistic from a deployment perspective. Reducing the complexity of mobile node to network security for localized mobility management can therefore reduce barriers to deployment. Removing mobile node involvement in localized mobility management also limits the possibility of DoS attacks on network infrastructural elements. In a host based approach, the mobile node is required to have a global or restricted routing local IP address for a network infrastructure element, the mobility anchor point. The network infrastructural element therefore becomes a possible target for DoS attacks, even if mobile nodes are properly authenticated. A properly authenticated mobile node can either willfully or inadvertently give the network infrastructural element address to an attacker. In summary, ruling out mobile node involvement in local mobility management simplifies security by removing the need for service-specific credentials to authenticate and authorize the mobile node for localized mobility management in the network and by limiting the possibility of DoS attacks on network infrastructural elements. The goal is that support for localized mobility management should not require additional security between the mobile node and the network. 2.7 Support for Heterogeneous Wireless Link Technologies (Goal #7) The number of wireless link technologies available is growing, and the growth seems unlikely to slow down. Since the standardization of a wireless link PHY and MAC is a time consuming process, reducing the amount of work Kempf, et. al. Expires October 2006 [Page 6] Internet Draft LMM Goals and Gap Analysis April, 2006 necessary to interface a particular wireless link technology to an IP network is necessary. A localized mobility management solution should ideally require minimal work to interface with a new wireless link technology. In addition, an edge mobility solution should provide support for multiple wireless link technologies within the network in separate subnets. It is not required that the localized mobility management solution support handover from one wireless link technology to another without change in IP address. The reason is because a change in network interface typically requires that the hardware interface associated with one wireless link technology be brought up and configured, and this process typically requires that the IP stack for the new interface card be configured on the mobile node from the driver up. Requiring that the mobile node IP stack circumvent this process to keep the IP address constant would be a major change in the way the IP stack software is implemented. The goal is that the localized mobility management protocol should not use any wireless link specific information for basic routing management, though it may be used for other purposes, such as identifying a mobile node. 2.8 Support for Unmodified Mobile Nodes (Goal #8) In the wireless LAN switching market, no modification of the software on the mobile node is required to achieve "IP mobility" (which means localized mobility management). Being able to accommodate unmodified mobile nodes enables a service provider to offer service to as many customers as possible, the only constraint being that the customer is authorized for network access. Another advantage of minimizing mobile node software for localized mobility management is that multiple global mobility management protocols can be supported with a localized mobility management solution that does not have mobile node involvement. While Mobile IPv6 is clearly the global mobility management protocol of primary interest going forward, there are a variety of global mobility management protocols that might also need support, including proprietary protocols needing support for backward compatibility reasons. Within IETF, both HIP and Mobike are likely to need support in addition to Mobile IPv6, and Mobile IPv4 support may also be necessary. Note that this goal does NOT mean that the mobile node has no software at all associated with mobility or wireless. The mobile node must have some kind of global mobility protocol if it is to move from one domain of edge mobility support to another, although no global mobility protocol is required if the mobile node only moves within the coverage area of the localized mobility management protocol. Also, every wireless link protocol requires handover support on the mobile node in the physical and MAC layers, typically in the wireless interface driver. Information passed from the MAC layer to the IP layer on the mobile node may be necessary to trigger IP signaling for IP link handover. Such movement detection support at the IP level may be required in order to determine whether the mobile node's Kempf, et. al. Expires October 2006 [Page 7] Internet Draft LMM Goals and Gap Analysis April, 2006 default router is still reachable after the move to a new access point has occurred at the MAC layer. Whether or not such support is required depends on whether the MAC layer can completely hide link movement from the IP layer. For a wireless link protocol such as the 3G protocols, the mobile node and network undergo an extensive negotiation at the MAC layer prior to handover, so it may be possible to trigger routing update without any IP protocol involvement. However, for a wireless link protocol such as IEEE 802.11 in which there is no network involvement in handover, IP layer movement detection signaling from the mobile node may be required to trigger routing update. The goal is that the localized mobility management solution should be able to support any mobile node that walks up to the link and has a wireless interface that can communicate with the network, without requiring localized mobility management software on the mobile node. 2.9 Support for IPv4 and IPv6 (Goal #9) While most of this document is written with IPv6 in mind, localized mobility management is a problem in IPv4 networks as well. A solution for localized mobility that works for both versions of IP is desirable, though the actual protocol may be slightly different due to the technical details of how each IP version works. From Goal #8 (Section 2.8), minimizing mobile node support for localized mobility means that ideally no IP version-specific changes would be required on the mobile node for localized mobility, and that global mobility protocols for both IPv4 and IPv6 should be supported. Any IP version-specific features would be confined to the network protocol. 2.10 Re-use of Existing Protocols Where Sensible (Goal #10) Many existing protocols are available as Internet Standards upon which the NETLMM protocol can be built. The design of the protocol should have a goal to re-use existing protocols where it makes architectural and engineering sense to do so. The design should not, however, attempt to re-use existing protocols where there is no real architectural or engineering reason. For example, the suite of Internet Standards contains several good candidate protocols for the transport layer, so there is no real need to develop a new transport protocol specifically for NETLMM. Re-use is clearly a good engineering decision in this case, since backward compatibility with existing protocol stacks is important. On the other hand, the network-based, localized mobility management functionality being introduced by NETLMM is a new piece of functionality, and therefore any decision about whether to re- use an existing global mobility management protocol should carefully consider whether re-using such a protocol really meets the needs of the functional architecture for network-based localized mobility management. The case for re-use is not so clear in this case, since there is no compelling backward compatibility argument. 3.0 Security Considerations Kempf, et. al. Expires October 2006 [Page 8] Internet Draft LMM Goals and Gap Analysis April, 2006 There are two kinds of security issues involved in network-based localized mobility management: security between the mobile node and the network, and security between network elements that participate in the network-based localized mobility management protocol Security between the mobile node and the network itself consists of two parts: threats involved in localized mobility management in general, and threats to the network-based localized mobility management from the host. The first is discussed above in Sections 2.3 and 2.6. The second is discussed in the threat analysis document [28]. For threats to network-based localized mobility management, the basic threat is an attempt by an unauthorized party to signal a bogus mobility event. Such an event must be detectable. This requires proper bidirectional authentication and authorization of network elements that participate in the network-based localized mobility management protocol, and data origin authentication on the signaling traffic between network elements. 4.0 Author Information James Kempf DoCoMo USA Labs 181 Metro Drive, Suite 300 San Jose, CA 95110 USA Phone: +1 408 451 4711 Email: kempf@docomolabs-usa.com Kent Leung Cisco Systems, Inc. 170 West Tasman Drive San Jose, CA 95134 USA EMail: kleung@cisco.com Phil Roberts Motorola Labs Schaumberg, IL USA Email: phil.roberts@motorola.com Katsutoshi Nishida NTT DoCoMo Inc. 3-5 Hikarino-oka, Yokosuka-shi Kanagawa, Japan Phone: +81 46 840 3545 Email: nishidak@nttdocomo.co.jp Gerardo Giaretta Telecom Italia Lab Kempf, et. al. Expires October 2006 [Page 9] Internet Draft LMM Goals and Gap Analysis April, 2006 via G. Reiss Romoli, 274 10148 Torino Italy Phone: +39 011 2286904 Email: gerardo.giaretta@tilab.com Marco Liebsch NEC Network Laboratories Kurfuersten-Anlage 36 69115 Heidelberg Germany Phone: +49 6221-90511-46 Email: marco.liebsch@ccrle.nec.de 5.0 Informative References [1] Kempf, J., Leung, K., Roberts, P., Nishda, K., Giaretta, G., Liebsch, M., and Gwon, Y., "Problem Statement for IP Local Mobility," Internet Draft, work in progress. [2] Vogt, C., and Kempf, J., "Security Threats to Network-based Localized Mobillity Management", Internet draft, work in progress. [3] Manner, J., and Kojo, M., "Mobility Related Terminology", RFC 3753, June, 2004. [4] Devarapalli,V., Wakikawa, R., Petrescu, A., Thubert, P., "Network Mobility (NEMO) Basic Support Protocol", RFC 3963, January, 2005. [5] Carpenter, B., "Architectural Principles of the Internet," RFC 1958, June, 1996. [6] Johnson, D., Perkins, C., and Arkko, J., "Mobility Support in IPv6", RFC 3775. [7] Moskowitz, R., Nikander, P., Jokela, P., and Henderson, T., "Host Identity Protocol", Internet Draft, work in progress. [8] Choi, J, and Daley, G., "Goals of Detecting Network Attachment in IPv6", Internet Draft, work in progress. [9] IEEE, "Port-based Access Control", IEEE LAN/MAN Standard 802.1x, June, 2001. [10] Forsberg, D., Ohba, Y., Patil, B., Tschofenig, H., and Yegin, A., "Protocol for Carrying Authentication for Network Access (PANA)", Internet Draft, work in progress. [11] Arkko, J., Kempf, J., Zill, B., and Nikander, P., "SEcure Neighbor Discovery (SEND)", RFC 3971, March, 2005. [12] Moore, N., "Optimistic Neighbor Discovery", Internet Draft, Work in Progress. [13] Ackerman, L., Kempf, J., and Miki, T., "Wireless Location Privacy: Law and Policy in the US, EU, and Japan", ISOC Member Briefing #15, http://www.isoc.org/briefings/015/index.shtml [14] Haddad, W., Nordmark, E., Dupont, F., Bagnulo, M., Park, S.D., and Patil, B., "Privacy for Mobile and Multi-homed Nodes: MoMiPriv Problem Statement", Internet Draft, work in progress. [15] Kivinen, T., and Tschopfening, H., "Design of the MOBIKE Protocol", Internet Draft, work in progress. [16] Koodli, R., "IP Address Location Privacy and IP Mobility", Internet Draft, work in progress. Kempf, et. al. Expires October 2006 [Page 10] Internet Draft LMM Goals and Gap Analysis April, 2006 [17] Koodli, R., Devarapalli, V., Flinck, H., and Perkins, C., "Solutions for IP Address Location Privacy in the presence of IP Mobility", Internet Draft, work in progress. [18] Bao, F., Deng, R., Kempf, J., Qui, Y., and Zhou, J., "Protocol for Protecting Movement of Mobile Nodes in Mobile IPv6", Internet Draft, work in progress. [19] Soliman, H., Tsirtsis, G., Devarapalli, V., Kempf, J., Levkowetz, H., Thubert, P, and Wakikawa, R. "Dual Stack Mobile IPv6 (DSMIPv6) for Hosts and Routers", Internet Draft, work in progress. [20] Koodli, R., editor, "Fast Handovers for Mobile IPv6", RFC 4068, July, 2005. [21] Soliman, H., Castelluccia, C., El Malki, K., and Bellier, L., "Hierarchical Mobile IPv6 Mobility Management (HMIPv6)", RFC 4140, August, 2005. [22] Kempf, J., and Koodli, R., "Bootstrapping a Symmetric IPv6 Handover Key from SEND", Internet Draft, work in progress. [23] Campbell, A., Gomez, J., Kim, S., Valko, A., and Wan, C., "Design, Implementation and Evaluation of Cellular IP", IEEE Personal Communications, June/July 2000. [24] Ramjee, R., La Porta, T., Thuel, S., and Varadhan, K., "HAWAII: A domain-based approach for supporting mobility in wide-area wireless networks", in Proceedings of the International Conference on Networking Protocols (ICNP), 1999. [25] "Mobile VPN Network Configuration Alternatives", IP Unplugged, http://www.ipunplugged.com/pdf/Network-blueprints_A.pdf. [26] Oran, D., "OSI IS-IS Intra-domain Routing Protocol", RFC 1142, Feburary, 1990. [27] Moy, J., "OSPF Version 2", STD 54, April, 1998. [28] Threat analysis draft, TBD 6.0 IPR Statements The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org. Kempf, et. al. Expires October 2006 [Page 11] Internet Draft LMM Goals and Gap Analysis April, 2006 7.0 Disclaimer of Validity This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. 8.0 Copyright Notice Copyright (C) The Internet Society (2006). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. 9.0 Appendix: Gap Analysis This section discusses a gap analysis between existing proposals for solving localized mobility management and the goals in Section. 2.0. 9.1 Mobile IPv6 with Local Home Agent One option is to deploy Mobile IPv6 with a locally assigned home agent in the local network. This solution requires the mobile node to somehow be assigned a home agent in the local network when it boots up. This home agent is used instead of the home agent in the home network. The advantage of this option is that no special solution is required for edge mobility - the mobile node reuses the global mobility management protocol for that purpose - if the mobile node is using Mobile IPv6. One disadvantage is that Mobile IP has no provision for handover between home agents. Although such handover should be infrequent, it could be quite lengthy and complex. The analysis of this approach against the goals above is the following. Goal #1: If the mobile node does not perform route optimization, this solution reduces, but does not eliminate, IP link handover related performance problems. Goal #2: Similarly to Goal #1, signaling volume is reduced if no route optimization signaling is done on handover. Goal #3: Location privacy is preserved for external correspondents, but the mobile node itself still maintains a local care-of address which a worm or other exploit could misuse. If the mobile node does perform route optimization, location privacy may be compromised, and this solution is no better than having a remote home agent. Goal #4: If traffic is not route optimized, the mobile node still pays for an over-the-air tunnel to the locally assigned home agent. The overhead here Kempf, et. al. Expires October 2006 [Page 12] Internet Draft LMM Goals and Gap Analysis April, 2006 is exactly the same as if the mobile node's home agent in the home network is used and route optimization is not done. Goal #5: If the localized mobility management domain is large, the mobile node may suffer from unoptimzed routes. RFC 3775 [6] provides no support for notifying a mobile node that another localized home agent is available for a more optimized route, or for handing over between home agents. A mobile node would have to perform the full home agent bootstrap procedure, including establishing a new IPsec SA with the new home agent. Goal #6: A local home agent in a roaming situation requires the guest mobile node to have the proper credentials to authenticate with the local home agent in the serving network. Although the credentials used for network access authentication could also be used for mobile service authentication and authorization if the local home agent uses EAP over IKEv2 to authenticate the mobile node with its home AAA server, the two sets of credentials are in principle different, and could require additional credential provisioning. In addition, as in Goal #3, if binding updates are done in cleartext over the access network or the mobile node has become infected with malware, the local home agent's address could be revealed to attackers and the local home agent could become the target of a DoS attack. So a local home agent would provide no benefit for this goal. Goal #7: This solution supports multiple wireless technologies in separate IP link subnets. No special work is required to interface a local home agent to different wireless technologies. Goal #8: The mobile node must support Mobile IPv6 in order for this option to work. So mobile node changes are required and other IP mobility protocols are not supported. Goal #9: The Mobile IPv6 working group is working on modifying the protocol to allow registration of IPv4 care-of addresses in an IPv6 home agent, and also to allow a mobile node to have an IPv4 care of address [19]. Goal #10 This solution re-uses an existing protocol, Mobile IPv6. 9.2 Hierarchical Mobile IPv6 (HMIPv6) HMIPv6 [21] provides the most complete localized mobility management solution available today as an RFC. In HMIPv6, a localized mobility anchor called a MAP serves as a routing anchor for a regional care-of address. When a mobile node moves from one access router to another, the mobile node changes the binding between its regional care-of address and local care-of address at the MAP. No global mobility management signaling is required, since the care-of address seen by correspondents does not change. This part of HMIPv6 is similar to the solution outlined in Section 9.1; however, HMIPv6 also allows a mobile node to hand over between MAPs. Handover between MAPs and MAP discovery requires configuration on the routers. MAP addresses are advertised by access routers. Handover happens by overlapping MAP coverage areas so that, for some number of access routers, Kempf, et. al. Expires October 2006 [Page 13] Internet Draft LMM Goals and Gap Analysis April, 2006 more than one MAP may be advertised. Mobile nodes need to switch MAPs in the transition area, and then must perform global mobility management update and route optimization to the new regional care-of address, if appropriate. The analysis of this approach against the goals above is the following. Goal #1 This solution shortens, but does not eliminate, the latency associated with IP link handover, since it reduces the amount of signaling and the length of the signaling paths. Goal #2 Signaling volume is reduced simply because no route optimization signaling is done on handover within the coverage area of the MAP. Goal #3 Location privacy is preserved for external correspondents, but the mobile node itself still maintains a local care-of address which a worm or other exploit could access by sending the local care-of address to third malicious node to enable it to track the mobile node's location. Goal #4 The mobile node always pays for an over-the-air tunnel to the MAP. If the mobile node is tunneling through a global home agent or VPN gateway, the wired link experiences double tunneling. Over-the-air tunnel overhead can be removed by header compression, however. Goal #5 From Goal #1 and Goal #4, the signaling overhead is no more or less than for mobile nodes whose global mobility management anchor is local. However, because MAP handover is possible, routes across large localized mobility management domains can be improved thereby improving wired network resource utilization by using multiple MAPs and handing over, at the expense of the configuration and management overhead involved in maintaining multiple MAP coverage areas. Goal #6 In a roaming situation, the guest mobile node must have the proper credentials to authenticate with the MAP in the serving network. In addition, since the mobile node is required to have a unicast address for the MAP that is either globally routed or routing restricted to the local administrative domain, the MAP is potentially a target for DoS attacks across a wide swath of network topology. Goal #7 This solution supports multiple wireless technologies in separate IP link subnets. Goal #8 This solution requires modification to the mobile nodes. In addition, the HMIPv6 design has been optimized for Mobile IPv6 mobile nodes, and is not a good match for other global mobility management protocols. Goal #9 Currently, there is no IPv4 version of this protocol; although there is an expired Internet draft with a design for a regional registration protocol for Mobile IPv4 that has similar functionality. It is possible that the same IPv4 transition solution as used for Mobile IPv6 could be used [19]. Goal #10 This solution re-uses an existing protocol, HMIPv6. Kempf, et. al. Expires October 2006 [Page 14] Internet Draft LMM Goals and Gap Analysis April, 2006 9.3 Combinations of Mobile IPv6 with Optimizations One approach to local mobility that has received much attention in the past and has been thought to provide a solution is combinations of protocols. The general approach is to try to cover gaps in the solution provided by MIPv6 by using other protocols. In this section, gap analyses for MIPv6 + FMIPv6 and HMIPv6 + FMIPv6 are discussed. 9.3.1 MIPv6 + FMIPv6 As discussed in Section 9.1, the use of MIPv6 with a dynamically assigned, local home agent cannot fulfill the goals. A fundamental limitation is that Mobile IPv6 cannot provide seamless handover (i.e. Goal #1). FMIPv6 has been defined with the intent to improve the handover performance of MIPv6. For this reason, the combined usage of FMIPv6 and MIPv6 with a dynamically assigned local home agent has been proposed to handle local mobility. Note that this gap analysis only applies to localized mobility management, and it is possible that MIPv6 and FMIPv6 might still be acceptable for global mobility management. The analysis of this combined approach against the goals follows. Goal #1 FMIPv6 provides a solution for handover performance improvement that should fulfill the goals raised by real-time applications in terms of jitter, delay and packet loss. The location of the home agent (in local or home domain) does not affect the handover latency. Goal #2 FMIPv6 requires the mobile node to perform extra signaling with the access router (i.e. exchange of RtSolPr/PrRtAdv and FBU/FBA). Moreover, as in standard MIPv6, whenever the mobile node moves to another IP link, it must send a Binding Update to the home agent. If route optimization is used, the mobile node also performs return routability and sends a Binding Update to each correspondent node. Nonetheless, it is worth noting that FMIPv6 should result in a reduction of the amount of IPv6 Neighbor Discovery signaling on the new link. Goal #3 The mobile node mantains a local care-of address. If route optimization is not used, location privacy can be achieved using bi- directional tunneling. However, as mentioned in Section 9.1, a worm or other malware can exploit this care of address by sending it to a third malicious node. Goal #4 As stated for Goal #2, the combination of MIPv6 and FMIPv6 generates extra signaling overhead. For data packets, in addition to the Mobile IPv6 over-the-air tunnel, there is a further level of tunneling between the mobile node and the previous access router during handover. This tunnel is needed to forward incoming packets to the mobile node addressed to the previous care-of address. Another reason is that, even if the mobile node has a valid new care-of address, the mobile node cannot use the new care of address directly with its correspondents without performing route Kempf, et. al. Expires October 2006 [Page 15] Internet Draft LMM Goals and Gap Analysis April, 2006 optimization to the new care of address. This implies that the transient tunnel overhead is in place even for route optimized traffic. Goal #5 FMIPv6 generates extra signaling overhead between the previous access router and the new access router for the HI/HAck exchange. Concerning data packets, the use of FMIPv6 for handover performance improvement implies a tunnel between the previous access router and the mobile node that adds some overhead in the wired network. This overhead has more impact on star topology deployments, since packets are routed down to the old access router, then back up to the aggregation router and then back down to the new access router. Goal #6 In addition to the analysis for Mobile IPv6 with local home agent in Section 9.1, FMIPv6 requires the mobile node and the previous access router to share a security association in order to secure FBU/FBA exchange. So far, only a SEND-based solution has been proposed and this requires the mobile node to use autoconfigured Cryptographically Generated Addresses (CGAs)[22]. This precludes stateful address allocation using DHCP, which might be a necessary deployment in certain circumstances. Another solution based on AAA is under study but it could require extra signaling overhead over the air and in the wired network and it could raise performance issues. Goal #7 MIPv6 and FMIPv6 support multiple wireless technologies, so this goal is fufilled. Goal #8 The mobile node must support both MIPv6 and FMIPv6, so it is not possible to satisfy this goal. Goal #9 Work is underway to extend MIPv6 with the capability to run over both IPv6-enabled and IPv4-only networks [19]. FMIPv6 only supports IPv6. Even though an IPv4 version of FMIP has been recently proposed, it is not clear how it could be used together with FMIPv6 in order to handle fast handovers across any wired network. Goal #10 This solution re-uses existing protocols, Mobile IPv6 and FMIPv6. 9.3.2 HMIPv6 + FMIPv6 HMIPv6 provides several advantages in terms of local mobility management. However, as seen in Section 9.2, it does not fulfill all the goals identified in Section 2.0. In particular, HMIPv6 does not completely eliminate the IP link handover latency. For this reason, FMIPv6 could be used together with HMIPv6 in order to cover the gap. Note that even if this solution is used, the mobile node is likely to need MIPv6 for global mobility management, in contrast with the MIPv6 with dynamically assigned local home agent + FMIPv6 solution. Thus, this solution should really be considered MIPv6 + HMIPv6 + FMIPv6. The analysis of this combined approach against the goals follows. Kempf, et. al. Expires October 2006 [Page 16] Internet Draft LMM Goals and Gap Analysis April, 2006 Goal #1 HMIPv6 and FMIPv6 both shorten the latency associated with IP link handovers. In particular, FMIPv6 is expected to fulfill the goals on jitter, delay and packet loss raised by real-time applications. Goal #2 Both FMIPv6 and HMIPv6 require extra signaling compared with Mobile IPv6. As a whole the mobile node performs signaling message exchanges at each handover that are RtSolPr/PrRtAdv, FBU/FBA, LBU/LBA and BU/BA. However, as mentioned in Section 9.2, the use of HMIPv6 reduces the signaling overhead since no route optimization signaling is done for intra-MAP handovers. In addition, naïve combinations of FMIPv6 and HMIPv6 often result in redundant signaling. There is much work in the academic literature and the IETF on more refined ways of combining signaling from the two protocols to avoid redundant signaling. Goal #3 HMIPv6 may preserve location privacy, depending on the dimension of the geographic area covered by the MAP. As discussed in Section 9.2, the mobile node still maintains a local care-of address that can be exploited by worms or other malware. Goal #4 As mentioned for Goal #2, the combination of HMIPv6 and FMIPv6 generates a lot of signaling overhead in the network. Concerning payload data, in addition to the over-the-air MIPv6 tunnel, a further level of tunneling is established between mobile node and MAP. Notice that this tunnel is in place even for route optimized traffic. Moreover, if FMIPv6 is directly applied to HMIPv6 networks, there is a third temporary handover- related tunnel between the mobile node and previous access router. Again, there is much work in the academic literature and IETF on ways to reduce the extra tunnel overhead on handover by combining HMIP and FMIP tunneling. Goal #5 The signaling overhead in the network is not reduced by HMIPv6, as mentioned in Section 9.2. Instead, FMIPv6 generates extra signaling overhead between the previous access router and new access router for HI/HAck exchange. For payload data, the same considerations as for Goal #4 are applicable. Goal #6 FMIPv6 requires the mobile node and the previous access router to share a security association in order to secure the FBU/FBA exchange. In addition, HMIPv6 requires that the mobile node and MAP share an IPsec security association in order to secure LBU/LBA exchange. The only well understood approach to set up an IPsec security association using of certificates, but this may raise deployment issues. Thus, the combination of FMIPv6 and HMIPv6 doubles the amount of mobile node to network security protocol required, since security for both FMIP and HMIP must be deployed. Goal #7 HMIPv6 and FMIPv6 support multiple wireless technologies, so this goal is fufilled. Goal #8 The mobile node must support both HMIPv6 and FMIPv6 protocols, so this goal is not fulfilled. Kempf, et. al. Expires October 2006 [Page 17] Internet Draft LMM Goals and Gap Analysis April, 2006 Goal #9 Currently there is no IPv4 version of HMIPv6. There is an IPv4 version of FMIP but it is not clear how it could be used together with FMIPv6 in order to handle fast handovers across any wired network. Goal #10 This solution re-uses existing protocols, HMIPv6 and FMIPv6. 9.4 Micromobility Protocols Researchers have defined some protocols that are often characterized as micromobility protocols. Two typical protocols in this category are Cellular-IP [23] and HAWAII [24]. Researchers defined these protocols before local mobility optimizations for Mobile IP such as FMIP and HMIP were developed, in order to reduce handover latency. The micromobility approach to localized mobility management requires host route propagation from the mobile node to a collection of specialized routers in the localized mobility management domain along a path back to a boundary router at the edge of the localized mobility management domain. A boundary router is a kind of localized mobility management domain gateway. Localized mobility management is authorized with the access router, but reauthorization with each new access router is necessary on IP link movement, in addition to any reauthorization for basic network access. The host routes allow the mobile node to maintain the same IP address when it moves to a new IP link, and still continue to receive packets on the new IP link. Cellular IP and HAWAII have a few things in common. Both are compatible with Mobile IP and are intended to provide a higher level of handover performance in local networks than was previously available with Mobile IP without such extensions as HMIP and FMIP. Both use host routes installed in a number of routers within a restricted routing domain. Both define specific messaging to update those routes along the forwarding path and specify how the messaging is to be used to update the routing tables and forwarding tables along the path between the mobile and a micromobility domain boundary router at which point Mobile IP is to used to handle scalable global mobility. Unlike the FMIP and HMIP protocols, however, these protocols do not require the mobile node to obtain a new care of address on each access router as it moves; but rather, the mobile node maintains the same care of address across the micromobility domain. From outside the micromobility domain, the care of address is routed using traditional longest prefix matching IP routing to the domain's boundary router, so the care of address conceptually is within the micromobiity domain boundary router's subnet. Within the micromobility domain, the care of address is routed to the correct access router using host routes. Cellular IP and HAWAII differ in a few aspects. Cellular IP seems to be restricted, based on the nature of the protocol, to tree-based network topologies. HAWAII claims to be applicable in both tree-based and more complete network topologies. HAWAII documents more functionality in the realm of reliability and QoS interworking. Both protocols involve the mobile node itself in mobility operations, although this is also true of the Mobile IP based optimizations, so both protocols raise the same security Kempf, et. al. Expires October 2006 [Page 18] Internet Draft LMM Goals and Gap Analysis April, 2006 concerns with respect to authorizing address update as the Mobile IP based optimizations. HAWAII provides some analysis of network deployment scenarios including scale, density, and efficiency, but some of these assumptions seem very conservative compared to realistic cellular type deployments. Micromobility protocols have some potential drawbacks from a deployment and scalability standpoint. Both protocols involve every routing element between the mobile device and the micromobility domain boundary router in all packet forwarding decisions specific to care of addresses for mobile nodes. Scalability is limited because each care of address corresponding to a mobile node generates a routing table entry, and perhaps multiple forwarding table entries, in every router along the path. Since mobile nodes can have multiple global care of addresses in IPv6, this can result in a large expansion in router state throughout the micromobility routing domain. Although the extent of the scalability for micromobility protocols is still not clearly understood from a research standpoint, it seems certain that they will be less scalable than the Mobile IP optimization enhancements, and will require more specialized gear in the wired network. The following is a gap analysis of the micromobility protocols against the goals in Section 2.0: Goal #1 The micromobility protocols reduce handover latency by quickly fixing up routes between the boundary router and the access router. While some additional latency may be expected during host route propagation, it is typically much less than experienced with standard Mobile IP. Goal #2 The micromobility protocols require signaling from the mobile node to the access router to initiate the host route propagation, but that is a considerable reduction over the amount of signaling required to configure to a new IP link. Goal #3 No care-of address changes are exposed to correspondent nodes or the mobile node itself while the mobile node is moving in the micromobility- managed network. Because there is no local care-of address, there is no threat from malware that exposes the location by sending the care-of address to an adversary. Goal #4 The only additional over-the-air signaling is involved in propagating host routes from the mobile node to the network upon movement. Since this signaling would be required for movement detection in any case, it is expected to be minimal. Mobile node traffic experiences no overhead. Goal #5 Host route propagation is required throughout the wired network. The volume of signaling could be more or less depending on the speed of mobile node movement and the size of the wired network. Goal #6 The mobile node only requires a security association of some type with the access router. Because the signaling is causing routes to the mobile node's care-of address to change, the signaling must prove authorization to hold the address. Kempf, et. al. Expires October 2006 [Page 19] Internet Draft LMM Goals and Gap Analysis April, 2006 Goal #7 The micromobility protocols support multiple wireless technologies, so this goal is satisfied. Goal #8 The mobile node must support some way of signaling the access router on link handover, but this is required for movement detection anyway. The nature of the signaling for the micromobility protocols may mobile node software changes, however. Goal #9 Most of the work on the micromobility protocols was done in IPv4, but little difference could be expected for IPv6. Goal #10 This solution does not reuse an existing protocol because there is currently no Internet Standard protocol for micromobility. 9.5 Standard Internal Gateway Route Distribution Protocols (OSPF and IS-IS) It has also been suggested that instead of using a specialized micromobility routing protocol in the access network, a standardized Internal Gateway route distribution Protocol (IGP) such as IS-IS [26] or OSPF [27] should be used to distribute host routes. Host route messages are formatted in the IGPs by including a subnet mask in the route information update, indicating that the address is a /32 for IPv4 or a /128 for IPv6 instead of a subnet prefix. Since IGPs typically distribute route information by flooding, every router in the domain obtains a copy of the host route eventually. Using an IGP instead of a micromobility protocol has the advantage that standardized routing equipment can be used for all routers in the access network, and, if a particular router goes down, the host routes maintained along alternate routes should be sufficient to continue routing, unlike micromobility protocols where only targeted routers have the host routes. Distributing host routes with an IGP has some significant disadvantages however. One is that flooding requires a certain amount of time to converge; so for some period after the link handover blackout, delivery to a mobile node that has moved will be disrupted until convergence along the routes traveled by the mobile node's traffic has occurred. Because micromobility protocols target host routes back to the micromobility domain border router, convergence can potentially be achieved more quickly. Tunnel-based solutions such as HMIP don't suffer from convergence latency because only the two endpoints need to know the routing. As a result, the internal routing tables updated by the IGP remain stable when a mobile node moves. The movement does not affect routing of traffic to other mobile nodes due to intensive path computation. Another disadvantage of using an IGP is that each router in the domain must maintain a full host route table for all hosts. This goal raises a scalability issue. For example, an experiment [25] using OSPF to distribute host routes through an access network consisting of a collection of rather smallish enterprise routers indicated that about 25,000 routes could be supported in 22 Mb of memory before performance started to degrade. This works out to about 0.88 kb/host. Scaling this up to, say, 10 million hosts Kempf, et. al. Expires October 2006 [Page 20] Internet Draft LMM Goals and Gap Analysis April, 2006 (what one might expect in a large metropolitan area such as Tokyo or San Francisco) would require about 8.8 Gb of memory per router. While memory costs continue to drop, the amount of processing power for searching a routing database of that size essentially means that each router in the domain must have the same processing power as a high end, border router. Micromobility and tunnel- based solutions don't suffer from this problem, because the host route is local to the tunnel endpoints. Other routers in the domain route based on highly scalable shortest matching network prefix method. The following is a gap analysis of host route distribution using a standardized IGP against the goals in Section 2.0: Goal #1 Host route distribution using an IGP is likely to suffer from increased handover latency due to a lag time as routes converge over the access network. The exact amount of latency depends on the convergence characteristics of the particular IGP and the topology of the access network, i.e. how fast convergence occurs along routes taken by the mobile node's traffic. Goal #2 Host route distribution using an IGP requires signaling from the mobile node to the access router to initiate the host route propagation, but that is a considerable reduction over the amount of signaling required to configure to a new IP link. However, if a mobile node is moving quickly, the flooding traffic necessary to propagate a host route may not converge before the mobile node hands over again. This could result in a cacscading series of host route updates that never converge. Whether or not this effect occurs depends on the size of the localized mobility domain, and so the need to ensure convergence places an upper bound on the size of the domain or expected movement speed of the mobile nodes. Goal #3 No care-of address changes are exposed to correspondent nodes or the mobile node itself while the mobile node is moving in the localized mobility management domain. Because there is no local care-of address, there is no threat from malware that exposes the location by sending the care-of address to an adversary. Goal #4 The only additional over-the-air signaling involved is signaling from the mobile node to the access router indicating that the mobile node has moved. Mobile node traffic experiences no overhead. Goal #5 Host route propagation is required throughout the wired network. The volume of signaling could be more or less depending on the speed of mobile node movement and the size of the wired network. Goal #6 The mobile node only requires a security association of some type with the access router. Because the signaling is causing routes to the mobile node's care-of address to change, the signaling must prove authorization to hold the address. Goal #7 This goal is satisfied by default, since the IGPs are used over the wired backbone. Kempf, et. al. Expires October 2006 [Page 21] Internet Draft LMM Goals and Gap Analysis April, 2006 Goal #8 The mobile node needs to perform some kind of active movement detection with proof of identity to trigger the host route distribution, but this kind of signaling is needed for movement regardless of whether localized mobility management is deployed. Depending on the wireless link type, this may be handled by the wireless link layer. Goal #9 Since the IGPs support both IPv4 and IPv6, both are supported by this technique. Goal #10 This solution re-uses existing protocols, OSPF or IS-IS. 9.6 Summary The following table summarizes the discussion in Section 9.1 through 9.5. In the table, a "M" indicates that the protocol completely meets the goal, a "P" indicates that it partially meets the goal, and an "X" indicates that it does not meet the goal. Protocol #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 -------- -- -- -- -- -- -- -- -- -- --- MIPv6 P X X X X X M X M M HMIPv6 P X X X P X M X X M MIPv6 + FMIPv6 M X X X P X M X P M HMIPv6 + FMIPv6 M X X X X X M X X M Micro. M M M M P M M M P X IGP X M M M X M M M M M Kempf, et. al. Expires October 2006 [Page 22]