A Mobility Management Solution Based on ID/Locator Separation

This work is funded by the European Commission funded ICT-FP7 IP Project EFIPSANS under Grant No. INFSO-ICT-215549, and the National Basic Research Program of China (“973” Program) under Grant No. 2009CB320504.

    The basic TCP/IP protocols of the Internet are designed for static hosts, and an IP address bears the semantics of locator and identifier. So the Internet does not, by nature, support mobility. To address the mobility issue, the IETF has proposed several mobility management protocols based on IP. These include mobile IP (MIP) [1],[2] and proxy MIP version 6 (PMIPv6) [3]. These protocols enable the Internet to support moving hosts. But they cannot solve the problem of routing scalability caused by the double semantics of IP addresses [4]-[6].

    To solve these problems, the basic approach is to decouple the dual semantics of IP addresses. Some renowned institutes have conducted research on this issue and put forward solutions. These solutions include Host Identity Protocol (HIP) [7], by R. Moskowitz et al.; SHIM6 [8] by Nordmark et al.; Locator/ID Separation Protocol (LISP) [9] by Cisco; Six/One [10] by Christian Vogt; and global locator, local locator, and identifier split (GLI-split) [11] by Michael Menth et al. On top of the network architecture based on ID/locator separation, traditional mobility management protocols have poor scalability. The centralized deployment of mobility management entities makes these entities prone to single point failure. Moreover, they are likely to become bottlenecks for the data communications of mobile hosts. Therefore, traditional mobility management solutions are no longer suitable, and a new solution is needed for network architecture based on ID/locator separation.

    This paper analyzes the disadvantages of existing mobility management solutions based on
ID/locator separation. It proposes a novel network architecture and a new mobility management solution. This new solution is designed to address the mobility problem by overwriting the destination address.

1 Existing Mobility Management Solutions Based on ID/Locator Separation
    Existing mobility management solutions based on ID/locator separation fall into three categories: host-based, network-based, and host and network-based.

    HIP is a typical host-based solution. In HIP, a host identity layer is added between the transport layer and the IP layer to decouple the dual semantics of an IP address. The transport layer and its upper layers are shielded from any change of the host IP address. But deploying HIP is difficult because it requires changing the host and deploying a large number of RendezVous servers (RVSs) [12] in the network. Moreover, HIP does not support broadcast services. When the two communicating parties move at the same time, a long handover delay is incurred. The SHIM6 protocol is another host-based solution. It divides the IP layer into three sublayers. Among them, the SHIM sublayer maintains the association between the identifiers and the locators of the two hosts in each session. These sublayers also shield the upper layers from any change of the hosts’ locators. However, detecting available address pairs may introduce a long delay; and thus, it provides poor mobility support.

    LISP is a network-based solution that only requires enhancement of the functions of edge routers in edge networks. No modification of the host is required. However, LISP is a protocol based on encapsulation. The data packets from a host must be encapsulated by the ingress tunnel router (ITR) before they are sent to the core network, which increases the bandwidth consumption. In LISP, each host is assigned a unique endpoint identifier (EID) in each edge network. When a host moves from one edge network to another during communication, the change of EID interrupts the TCP connection. To address this problem, LISP mobile node (LISP MN) architecture [13] is proposed. However, this architecture also has also some disadvantages. First, before deploying LISP MNs, functions of the hosts have to be enhanced. Second, an LISP MN can directly access the mapping system, which may cause some security problems for the entire system. Third, to enable interconnection between LISP MNs and traditional hosts, a proxy egress tunnel router (PETR) is needed in the network. Fourth, a mobile host does not have a mobile anchor point in the network. So data packets sent from the correspondent host (CH) to the mobile host in the period between when the mobile host starts to move and when the CH obtains the new IP address of the mobile host will be lost. A Six/One router is a network-based solution that uses an address overwriting method. The edge network is connected to the core network via Six/One routers. A host is assigned two addresses: a unique endpoint address in the edge network and a transmission address for global routing. The two addresses are one-to-one mapped, but the transmission overhead is quite large. Communication between an enhanced host and a traditional host is likely to be interrupted when the enhanced host moves.

    GLI-split architecture is a solution based on both host and network. It divides an IP address into three address spaces: global address, local address, and identifier address. On the host side, a vertical address translation function is introduced. It translates the identifier address used in the transport layer into a local or global address. The host uses the identifier address to set up a communication association. However, the two-level mapping system in GLI-split may cause a large handover delay. In addition, the host is allowed to access the mapping system, which may cause some security problems.

    In brief, existing mobility management solutions based on ID/locator have potential security problems, are difficult to deploy, and may have long handover delay. Therefore, it is necessary to suggest a new network architecture and develop a mobility management mechanism suitable to the architecture.

2 A New Mobility Management Solution Using Destination Address Overwriting
    An ideal mobility management solution based on ID/locator separation should:

• be compatible with traditional Internet so that deployment is less difficult and modifications to the host and Internet are kept to a minimum
• allow only network entities to access the mapping system to ensure system security
• have the mapping systems deployed close to or within an edge network to shorten handover delay
• have the mapping systems that store mapping between the ID and the locator of a host deployed in a distributed way in order to provide robustness
• have the address space used by the edge network separate from that used by the core network to make the network more scalable
• support various types of applications, including unicast, broadcast, and multicast applications
• support hosts with multiple interfaces so that multihoming services can be delivered.

2.1 New Network Architecture Based On ID/Locator Separation
    Fig. 1 illustrates the new network architecture based on ID/locator separation. In this architecture, the network is divided into core area (CA) and routing area (RA). The CA has the same functions as those of existing Internet backbone and consists of high-speed routers. One RA is made up of several organization areas (OAs).

    RA is edge network, and its range depends on the specific deployment. For example, an RA may be constructed according to geographical location. An OA is often related to an organization—one company can be an OA. Each RA is connected to the CA by one or several RA routers.

    The source and destination IP addresses of a packet are the IDs of the source and destination host respectively. The default gateway of the source host is configured as the address of the OA router the host is currently connected to. When an OA router receives a packet, it looks for the mapping between the ID and the locator in the local buffer based on the ID of the destination host. If a mapping record is found, the OA router forwards the packet within the OA. Otherwise, it uses the global routable IP address to overwrite the destination IP address field of the packet. First, it looks in the mapping system for the global routable address of the destination host according to the ID of the destination host and caches the result in the local buffer. Then, it uses this global routable address to overwrite the destination IP address field of the packet. After overwriting, the OA router sends the packet out. When the OA router of the destination host (denoted as OA-router-D) receives the packet, it looks up the ID of the destination host in the <ID, global locator> mapping stored locally according to the destination IP address. It then uses the found ID to overwrite the destination address field and forwards the packet to the destination host.

2.2 ID and Locator
    The ID is the unique information used to identify a host. So that the protocol stack and applications of the host are not modified, the IDs in the new network architecture are designed to have the same form as an IP address. The length of the ID is the same as an IPv6 address, that is, 128 bits. The ID is made up of two parts: RA information and host ID. The RA information is the prefix of the router that stores the host’s mapping information (denoted as Router-M). Its length is n bits. The host ID is the globally unique identity of the host. This ID can be generated, for example, in the same way as a host identifier tag (HIT) in HIP. However, only 128-n bits of the Hash value are truncated. The format of the ID is shown in Fig. 2.

    The solution proposed here is based on IPv6, so we assume that all traditional hosts support IPv6 protocol stack, and we call the host registered in the ID/locator separation-based system the extended host.

    Current solutions that use source and destination address overwriting have two problems in terms of interconnection between the extended host and the traditional host. One is that the global routable addresses of the extended hosts are stored in the domain name server (DNS). Because the DNS is updated at relatively long intervals, a traditional host may obtain the old IP address of an extended host. Consequently, the traditional host cannot initialize communication with the extended host. The other problem is that the session between an extended host and traditional host may be interrupted when the extended host moves and the traditional host fails to obtain the new global routable address of the extended host.

    Our solution to the first problem is to record the ID of the extended host in the DNS. Because the ID of the extended host is static, the connection establishment problem caused by slow DNS update can be avoided. Our solution to the second problem is to include some routing information in the ID and to adopt a communication mode with optimized routing based on proxy. Regardless of whether a session is initialized by an extended host or a traditional host, the source address of the first packet sent from the traditional host to the extended host is the global routable address of the traditional host; the destination address is the ID of the extended host. Because some routing information is included in the ID, the first packet will finally be routed to RA-Router-M. When the extended host moves from OA router 1 to OA router 2 during communication, OA router 1 is responsible for updating the location information of the extended host recorded at the correspondent host. If the correspondent host is an extended one, OA router 1 sends a Transfer message to the OA router of the correspondent host E (more details can be found in section 2.4). Otherwise, OA router 1 sends a normal IPv6 binding update message to the traditional host, which, in turn, processes the message according to MIPv6. Therefore, by including routing information in the ID and optimized proxy routing, our solution provides better backward compatibility.

    The locator is the global routing identity of a host, and it changes with the location of the host. It is a 128-bit IPv6 address consisting of RA, OA, and local locator. RA + OA is the prefix of the OA router to which the host connects. The local locator is a local address assigned by the OA router and is only valid within the coverage of the OA router. A host can only “see” its local locator. The format of the locator is shown in Fig. 3.

2.3 Mapping System
    The mapping system stores the relationship between the locator and the ID. It is comprised of local mapping systems within different RAs. To avoid introducing new entities to the network, the mapping system can be deployed on routers. To guarantee the robustness of the mapping system, the distributed hash table (DHT) is used to construct an overlay using routers within RAs.

    When a host is registered in the system for the first time, the local OA router stores the mapping between its ID and locator in the mapping system of the local RA. The router in which the mapping information is stored is referred to as Router-M. The selection of Router-M depends on the DHT protocol being used.

    When a host moves to another OA, the new OA router is responsible for updating the ID/locater mapping of the host in the mapping system. Because the RA field in the ID of the host contains the prefix of the RA where the Router-M is located, the new OA router can find the corresponding Router-M and update the ID/locater mapping of the host.

2.4 Mobility Management Solution
    The mobility management solution here involves location management and handover control. Three signaling processes are included: registration, location update, and handover control.

2.4.1 Registration Process
Registration occurs when a host joins the system for the first time. The local OA router registers the ID/locator mapping information of the host into the mapping system.
The OA router periodically broadcasts a router advertisement message. When a host enters the coverage of the OA router, it generates its own ID based on the router advertisement message and sends a Register message to the OA router. The parameter of the Register message is the ID of the host, ID-H. After receiving the Register message, the OA router assigns a local locator to the host and saves the mapping <ID-H, local locator> in local buffer. Then it performs a DHT operation and stores the mapping information of the host in the local mapping system of the local RA. Consequently, the host-related information, including the locators of both the host (global-locator-H) and the OA router (global-locator-OA-Router), is stored in Router-M. When the OA router receives the Put ACK message, it returns the Register ACK message to the host, indicating that registration is successful.

2.4.2 Location Update Process and
          Handover Control Process
Location update means the update of the mapping system on account of a change of the locator. This change is caused by movement of the host during communication with a CH. For optimized routing, the location update process should update information of the host in the mapping system as well as the host’s locator cached in the OA router of the CH.
A host may move in different scenarios. Here, we only take for example the most complicated case to describe the location update and handover control process. The host moves from the coverage of OA router 1 to the coverage of OA router 2 when it communicates with one CH. OA
router 1 and OA router 2 are in different RAs, denoted as old RA and new RA respectively.  Router-M of the host is in neither the old RA nor the new RA.
The location update and handover control process is as follows:
(1) The host moves into the coverage of OA router 2 and receives the advertisements broadcast by OA
router 2.
(2) After receiving the router advertisement message, the host knows that it has entered the coverage of a new OA router and sends a Register message containing its ID to OA router 2.
(3) After receiving the Register message from the host, OA router 2 assigns a local locator to the host and generates a global routable address (global-locator-H-new) for the host. Then it sends an Update message to RA router 2 requesting to update the location information of the host in the mapping system. In the update message, the host’s ID and new locator are included.
(4) When receiving the Update message, RA router 2 determines which RA the host’s Router-M is located in. RA router 2 determines this location based on the RA information in the
host’s ID and then sends an Update message to the RA router.
(5) Once the RA router (of the RA where the host’s Router-M is located) receives the Update message, it triggers the DHT update process within the local RA.
(6) Router-M then receives the Update message. It looks up the record of the host in the local buffer and sends a Forward message to OA router
1—where the information of the host has been updated with the new global routable address of the host and global routable address of OA router 2. The parameters of the Forward message include the host’s ID-H and new locator.
(7) After receiving the Forward message, OA router 1 stores the parameters of the message in local buffer. When it receives a data packet destined for the ID of the host, it uses the new global routable address of the host to overwrite the destination IP address field. Because OA router 1 has the mapping relation between the ID and the locator of the CH, it sends a Transfer message to the OA router that the CH connects to and requests to forward the data packet destined for the host to the new global routable address of the host.

2.5 Characteristics of the Mobile
      Management Solution
The mobility management solution proposed in this paper has some advantages. It is easy to deploy because the host’s protocol stack does not need to be modified, and no new network entities are needed. It is robust because the mapping information is organized using DHT, and it is secure because the host cannot access the mapping system directly.
The proposed solution has the following four characteristics:
(1) shortened distance between the router and the mapping system. This is because the mapping systems are deployed in different RAs in a distributed manner; and as a consequence, handover delay is decreased
(2) support for unicast, multicast, and broadcast without any modification of the multicasting protocols. The OA router records the IDs and locators of the hosts that join a multicast group. When it receives the multicast data, it delivers the data to the related hosts
(3) support for multi-interface hosts. Thus, multihoming schemes can be deployed. Each interface of a host can obtain a unique local locator from the local OA router. As a result, the host can receive data via multiple interfaces. The OA router maintains a connection table for the host, which contains the host’s ID, the CH’s ID, and the local locator of the host’s interface used to receive the data. When the OA router receives data destined for the host, it looks up the local locator in the connection table and forwards the data to the corresponding interface. Typically, a host can select the optimal interface based on factors such as access network status, user preference, network cost, and application type. It then updates the local locator saved in the OA router
(4) the address space used by the host is not included in the routing table of the core network. So no routing scalability problem will be caused.
3 Deployment of the
   Mobility Management
   Solution
Deployment of the suggested mobility management solution involves CA and RA functions. For instance, the CA and RAs can be deployed in an autonomous area. The functions of CA can be deployed on the backbone router of the autonomous area without any modification. The autonomous area can be divided into several RAs by geographical location. For example, a province can be an RA, and a city within a province can be an OA. On the egress routers of the RA and OA, the functions of RA router and OA router can be deployed. DHT protocol is deployed on these routers to form the mapping system.
4 Conclusion
This paper presents a new mobility management solution based on ID/Locator separation.  This solution uses destination address overwriting. It is compatible with the Internet and does not require any new entity to be deployed. The host cannot access the mapping system directly, so security for both users and the Internet is guaranteed. The mapping systems are distributed in different edge networks, and the distance for routers to access the mapping systems is shortened. Therefore, handover delay is also decreased. As a result, users have a better mobile experience. Multicast and broadcast are supported intrinsically as well as multi-interface hosts. Overall theoretical analysis and simulation tests of the system’s performance are our near future work.
T
Research Papers
A Mobility Management Solution Based on ID/Locator Separation
Yuhong Li, Yunjing Hou, and Shiduan Cheng
June 2011 Vol.9 No.2
ZTE COMMUNICATIONS
40
B
A
OA2
OA1
RA1_1 Router
RA1_2 Router
RA1
OA1 Router
ISP1
ISP2
RA2  Router
OA2 Router
OA1 Router
OA1
OA2
RA2
C
CA
▲Figure 1. A new network architecture based on ID/locator separation.
CA: core area
ISP: Internet service provider
OA: organization area
RA: routing area
OA2 Router
June 2011 Vol.9 No.2
ZTE COMMUNICATIONS
41
Research Papers
A Mobility Management Solution Based on ID/Locator Separation
Yuhong Li, Yunjing Hou, and Shiduan Cheng
Host ID: host identity                RA: routing area
RA
Host ID
n-bit
(128-n )-bit
▲Figure 2. Identifier format.
Research Papers
A Mobility Management Solution Based on ID/Locator Separation
Yuhong Li, Yunjing Hou, and Shiduan Cheng
June 2011 Vol.9 No.2
ZTE COMMUNICATIONS
42
RA: routing area              OA: organization area
RA
Local Locator
n-bit
64-bit
OA
(64-n )bit
▲Figure 3. Locator format.
June 2011 Vol.9 No.2
ZTE COMMUNICATIONS
43
Research Papers
A Mobility Management Solution Based on ID/Locator Separation
Yuhong Li, Yunjing Hou, and Shiduan Cheng
Biographies
Yuhong Li (hoyli@bupt.edu.cn) is an associate professor at the State Key Laboratory of Networking and Switching Technology of Beijing University of Posts & Telecommunications, China. Her research interests include mobility management, and next-generation Internet architecture.

Yunjing Hou (houyunjing@catt.cn) is a researcher at Datang Wireless Mobile Innovation Center, China Academy of Telecommunication Technology. Her research direction is mobility management and mobile network architecture.

Shiduan Cheng (chsd@bupt.edu.cn) is a professor and doctoral advisor at the State Key Laboratory of Networking and Switching Technology of Beijing University of Posts & Telecommunications, China. She has long been engaged in the research of telecommunication networks and computer networks.
References
[1] IETF RFC 5944 (2010, Nov.). IP Mobility Support for IPv4, Revised [Online]. Available: http://tools.ietf.org/html//rfc5944
[2] IETF RFC 3775 (2004, Jun.). Mobility Support in IPv6 [Online]. Available: http://www.ietf.org/rfc/rfc3775.txt
[3] IETF RFC 5213 (2008, Aug.). Proxy Mobile IPv6 [Online]. Available: http://tools.ietf.org/html/rfc5213
[4] IETF RFC 4984 (2007, Sept.). Report from the IAB Workshop on Routing and Addressing [Online]. Available: http://tools.ietf.org/html/rfc4984
[5] “BGP routing table analysis reports.” [Online]. Available: http://bgp.potaroo.net/
[6] G. Huston, “The BGP Instability Report.” [Online]. Available: http://bgpupdates.potaroo.net/instability/bgpupd.html
[7] IETF RFC 5201 (2008, Apr.). Host Identity Protocol [Online]. Available: http://tools.ietf.org/html//rfc5201
[8] IETF RFC 5533 (2009, Jun.). Shim6: Level 3 Multihoming Shim Protocol for IPv6 [Online]. Available: http://tools.ietf.org/html/rfc5533
[9] IETF (2010, Oct.). Locator/ID Separation Protocol (LISP) draft-ietf-lisp-09 [Online]. Available: http://tools.ietf.org/html/draft-ietf-lisp-09
[10] IETF (2009, Oct.). Six/One: A Solution for Routing and Addressing in IPv6 draft-vogt-rrg-six-one-02 [Online]. Available: http://tools.ietf.org/html/draft-vogt-rrg-six-one-02
[11] M. Menth, M. Hartmann, and D. Klein, “Global locator, local locator, and identifier split (GLI-split),” Inst. Comput. Sci., Uni. of Wurzburg, Germany, Tech. Rep. 470, Apr. 2010.
[12] IETF RFC 5204 (2008, Apr.). Host Identity Protocol (HIP) Rendezvous Extension [Online]. Available: http://tools.ietf.org/html/rfc5204
[13] IETF (2010, Oct.). LISP Mobile Node draft-meyer-lisp-mn-04 [Online]. Available: http://tools.ietf.org/html/draft-meyer-lisp-mn-04

    Taking into account these characteristics, this paper proposes a new network architecture based on ID/locator separation with a mobility management solution.

[Abstract] Current mobility management solutions based on ID/Locator separation are not easily deployed and cannot solve routing scalability and mobility problems. This paper proposes a novel network architecture based on ID/Locator separation and suggests a new mobility management solution. This solution solves the problem of scalability in the network and also provides better support for mobility. It can be easily deployed because no modification of the mobile host’s protocol stack is required. The identifier contains some routing information; so the solution provides intrinsic interworking with traditional mobile hosts. Because the mapping systems are distributed to the edge networks, robustness of the whole system is enhanced and handover delay is decreased.

[Keywords] identifier and locator separation; mobility management; location management; handoff control