Mobile IP Route Optimization

26.5.2000

Henrik Petander
Department of Electrical and Communications Engineering
Helsinki University of Technology
Henrik.Petander@hut.fi



 
 
 
 
 
 
 
 
 
 
 
 

Abstract

In standard IP routing routes are constantly optimized as routing tables are propagated throughout the Internet. In the case of mobile hosts that wish to use their home addresses this mechanism does not work. Thus while communicating sub-optimal routing often takes place. The problem results from the routing of packets to the mobile node via its home agent. Route optimization enables the datagrams to be routed directly in both directions. Route optimization also provides support for smooth handoffs by letting the previous foreign agent tunnel datagrams to mobile node's current location. Despite the advantages provided by route optimization it is unlikely that it will be deployed widely in the near future, due to its requirements for correspondent nodes.

Contents

1 Introduction

2 Addition of Mobility Support to IPv4
 2.1 Mobile IPv4 in general
 2.2 Triangle routing
 2.3 Bi-directional tunneling

3 Problems with Mobile IPv4
 3.1 Sub optimal routes
 3.2 Handoffs
 3.3 Criteria for an efficient solution

4 Mobile IPv4 route optimization
 4.1 Route optimization messages and data structures
 4.2 The effect on static routes
 4.3 Handoffs with route optimization
 4.4 Security considerations
 4.5 General deployment requirements

5 Mobile IPv6 and route optimization
 5.1 Main characteristics of Mobile IPv6
 5.2 The effect on routing
 5.3 The effect on handoffs
 5.4 Problems solved

 6 Conclusion

 Glossary

 References

 Further Information

1 Introduction

The introduction of portable computers and other mobile terminals with wireless network adapters has created a new need for mobility management in the Internet. In general users want to enjoy the same connectivity regardless of their physical location. They do not want to reboot their computers or restart their applications, even when they move from one network to another. To achieve this in the TCP/IP world, the addition of a mobility management protocol is needed. Some may argue that DHCP - Dynamic Host Configuration Protocol [2] or some other dynamic addressing protocol is a good enough solution. Even if these protocols do not require rebooting of the terminal they still break higher-level protocol connections.  To achieve true mobility while using the TCP/IP protocol suite, the IP protocol needs to be modified in a way that makes movement invisible to the higher level protocols.

Mobility in wireless networks can be implemented also in the link-layer. This makes it invisible to the network-layer and is often implemented with proprietary protocols. A link-layer solution is easier to deploy, since it requires changes only to the drivers of the network adapter whereas network layer solutions require broader changes to the operating system of the terminal. However, link-layer mobility is limited to a single subnetwork, which limits its applicability to large-scale user mobility.

In Section 2 I present the basic Mobile IPv4 protocol [9] in general and focus more deeply on its effect on routing. In Section 3 I discuss the problems related to Mobile IPv4 and criteria for solutions to these problems. In Section 4 I present route optimization and also analyze its performance and feasibility as a solution. In Section 5 I present Mobile IPv6 [4] and compare it to the route optimization in Mobile IPv4. In section 6 I  provide conclusions on route optimization based on the analysis in Sections 4 and 5.

2 Addition of mobility support to IPv4

IPv4 was not designed with later addition of mobility support in mind. The routing of datagrams is done with the network part of the IP-address. Thus all the addresses must correspond to the network topology in order for the nodes to be able to receive any datagrams from nodes in other subnetworks. Addition of mobility support to IPv4  has been done on top of  the IP-layer to minimize the required changes to hosts and routers. The signaling in mobile IPv4 is done with UDP datagrams.

2.1 Mobile IPv4 in general

Mobile IPv4 provides mobility support by allowing the mobile node to be reachable via its home address regardless of its physical location. This is achieved via tunneling of the datagrams to Mobile Node's current care-of address. When the Mobile Node moves to a different subnetwork, it sends a home registration to its home agent, which contains both the care-of address and the home address of the mobile node. With this information the home agent can capture and tunnel datagrams sent to the mobile node's home address.  The correspondent nodes do not need to be aware of the mobile node's location, as the mobility is invisible to them.

The care-of address of the mobile node acts as the tunnel endpoint. If the care-of address points to the mobile node it is called a co-located care-of address. The mobile node acquires the co-located care-of address via some mechanism, e.g., DHCP and sends a home registration request to its home agent, which can then start tunneling datagrams to the new address. In case the address points to a foreign agent it is called a foreign agent care-of address. Foreign agents may act as tunnel endpoints and provide mobility services for mobile nodes in foreign networks.

Upon entering the foreign network the mobile node first receives an agent advertisement, which contains the foreign agent care-of address. After this it sends a registration request to the foreign agent, which sends it further to the mobile node's home agent. Finally the home agent can tunnel packets to the foreign agent, which decapsulates them and forwards the original datagram to the mobile node. [9]

2.2 Triangle routing

Triangle routing is the basic routing scheme with Mobile IPv4. In triangle routing the mobile node sends its packets directly to the correspondent node. The correspondent node sends all datagrams to mobile node's home address. The home agent then tunnels them to mobile node's care-of address, as illustrated in figure 1. To preserve transport-layer connections mobile node uses its home address as the source address of all datagrams it sends.

Figure 1. Triangle routing with foreign agent care-of address.


 

2.3 Bi-directional tunneling

As the mobile node moves away from its home network it still continues to receive its packets via the home agent. In the case of bi-directional tunneling it also sends its packets via the home agent. The mobile node, or the foreign agent as in Figure 2., encapsulates the original datagram by adding a new header to it with the home agent's address as the destination address and the care-of address as the source address. Thus all traffic to and from the mobile node is routed via the home agent. This makes the mobility invisible to the correspondent nodes. However bi-directional tunneling with IP-IP tunneling adds 20 bytes of overhead to each packet sent by the mobile node, when compared to triangle routing. [8]

Figure 2. Bi-directional tunneling with foreign agent care-of address.

3 Problems with Mobile IPv4

3.1 Sub optimal routes

Both bi-directional tunneling and triangle routing lead to sub-optimal routes. Although triangle routing provides optimal routing from mobile node to correspondent node, it also leads to asymmetrical delays. The use of the home address as the source address in foreign networks is a questionable mechanism in the Internet of today, since routers performing ingress filtering will drop datagrams with a topologically incorrect source address [3]. In case the mobile node is on another continent than its home agent, the route from correspondent node to the mobile nodes via the home agent can cause long delays and also unnecessary network congestion.

Both schemes also consume extra bandwidth, due to the tunneling from the home agent to the care-of address that is used in both schemes. This can be partly alleviated with other tunneling techniques, which use smaller tunneling headers than IP-IP tunneling, such as minimal encapsulation [10]. Also header compression can be used to compress the inner header [1]. However, if a co-located care-of address is used, the tunneling is done also over the limited bandwidth radio medium between the network access point and the mobile node. With typical 572 byte IP datagrams this leads to a large overhead and wasted bandwidth.

3.2 Handoffs

Mobile IP was not designed for fast moving hosts. This is apparent in the movement detection algorithm in the specification, which contains two methods, which both are rather slow. The home agent handles all handoffs, although it may be far from the current network of the mobile node. The network delay adds to slow handoffs. Slow handoffs cause often packet loss, which is especially harmful  to real-time applications, such as voice over IP or video streaming. TCP-based connections also suffer, since lost packets may be mistaken for congestion and result in TCP's slow start mechanism [9].

Since the home agent handles handoffs, they cause lots of signaling traffic between the mobile node and the home agent. In high speed LANs this is not an issue, but when low speed WANs are involved and lots of mobile nodes are performing simultaneous handoffs, network congestion may result.

3.3 Criteria for an efficient solution

An efficient routing scheme would use direct routes between the mobile node and the correspondent node. It would also introduce minimal overhead for delivering the data and signaling information.  Handoffs would be performed with minimal packet loss. With localized handling of the handoffs can be performed somewhat faster, since the network delay between mobile node and home agent does not effect the handoff. Thus localization of handoffs would be a part of a good solution.

In addition to being technologically optimal, the solution, or protocol, should also be feasible. It should not require large changes to the operating systems of the correspondent nodes and  it should also be interoperable with the specifications of the TCP/IP protocol suite.

4 Mobile IPv4 route optimization

Mobile IPv4 route optimization [11] is a proposed extension to the Mobile IPv4 protocol. It provides enhancements to the routing of datagrams between the mobile node and to the correspondent node. The enhancements provide means for a correspondent node to tunnel datagrams directly to the mobile node or to its foreign agent care-of address.

4.1 Route optimization messages and data structures

The route optimization extension adds a conceptual data structure, the binding cache, to the correspondent node and to the foreign agent. The binding cache contains bindings for mobile nodes' home addresses and their current care-of addresses. With the binding the correspondent node can tunnel datagrams directly to the mobile node's care-of address.

Every time the home agent receives a datagram that is destined to a mobile node currently away from home, it sends a binding update to the correspondent node to update the information in the correspondent node's binding cache. After this the correspondent node can directly tunnel packets to the mobile node. Thus direct bi-directional communication is achieved with route optimization, as shown in Figure 3.

Figure 3. Direct routing with route optimization and foreign agent care-of address.

Route optimization adds four new UDP-messages to the Mobile IPv4 protocol:

Figure 4. Binding update to correspondent node

4.2 The effect on static routes

As the correspondent node learns the care-of address of the mobile node from the binding update, it can tunnel datagrams directly to the mobile node's care-of address [11]. Thus only the first datagrams are routed via the home agent. This reduces the network load and also reduces the delays caused by routing. Thus the optimization is valuable to mobile nodes that visit networks located far from their home agent.

However, the overhead caused by tunneling is not decreased. The correspondent node's use of minimal encapsulation [10] is a partial remedy, if both the encapsulator and the decapsulator support it. Ingress filtering [3] may also prevent the mobile node from sending datagrams directly to the correspondent node. The use of direct reverse tunneling from the care-of address to the correspondent node's address is a possible solution to ingress filtering. However, it is not possible with foreign agent care-of addresses, since the current reverse tunneling standard [8] requires the foreign agent to tunnel all packets to the home agent of the mobile node.

4.3 Smooth handoffs with route optimization

In the static case the protocol is fairly simple, but handoffs somewhat complicate the situation. When the correspondent node has an out of date entry for the mobile node's care-of address it tries to send the tunneled datagram to the mobile node's previous location and the datagram is lost. To solve this problem the protocol includes the previous foreign agent notification mechanism, which adds a binding cache to the foreign agent. [3]

When a mobile node moves to a new subnetwork it sends a registration request to the new foreign agent. The registration request may contain a previous foreign agent notification extension. Upon receiving such a request the foreign agent builds a binding update and sends it to the previous foreign agent. The previous foreign agent can then, after authenticating the update, create a binding for the mobile node. With this binding it can re-tunnel datagrams to the mobile node's new care-of address. The re-tunneling requires foreign agent care-of addresses in order for the agents to act as tunnel endpoints. [3]

The previous foreign agent notification mechanism provides temporary localization of the handoffs. It does not reduce the signaling load between the home agent and the mobile node, but reduces the number of datagrams lost due to correspondent nodes with out-of date bindings.

4.4 Security considerations

Since the correspondent nodes and foreign agents have binding caches, which change the routing of datagrams destined to mobile nodes, the binding updates must be authenticated. The authentication is performed in a similar manner as in base Mobile IPv4.  All binding updates contain a route optimization or smooth handoff authentication extension. This extension contains a hash, which is calculated from the datagram and the shared secret. [11]

The correspondent node and the mobile node's home agent need a security association [5]. This association is used for the authentication of the binding updates. Since the mobile node sends a binding update directly to its previous foreign agent, they also need a security association. If the security associations are not preconfigured they can be established via a key management protocol such as ISAKMP [6] or SKIP [7]. [11]

4.5 General deployment requirements

In order to make use of the binding updates the correspondent nodes must be able to process and authenticate them and be able to encapsulate datagrams [11]. To establish this the network stacks of the operating systems require changes. Since correspondent nodes need to establish a security association with the home agent and foreign agents need to establish one with the mobile node, a widely deployed key management system is obviously needed. Otherwise only nodes with statically configured security associations can benefit from the binding updates.

5 Mobile IPv6 and route optimization

5.1 Main characteristics of Mobile IPv6

Whereas Mobile IP was added on top of the IPv4 protocol, in IPv6 mobility support is built into the IP-layer [4]. In mobile IPv6 route optimization is an essential part of the protocol. Mobile nodes have a binding update list, which contains the bindings other nodes have for it. Correspondent nodes and home agents have a binding cache, which contains the home and care-of addresses of mobile nodes they have been recently communicating with. All signaling is performed via destination options that are appended to the base IPv6 header. Thus all signaling traffic can be piggybacked on datagrams with a data payload, as in Figure 5.

Figure 5. Destination option.

The destination options are:

All care-of addresses in Mobile IPv6 are co-located;  thus foreign agents are not a part of the protocol. Since all nodes are only required to understand the home address option, triangle routing will occur also with mobile IPv6. However, if the correspondent node implements the draft fully, only the first datagrams it sends will be routed via the home agent. The mobile node always sends a binding update to the original sender of a tunneled datagram. With this binding  the correspondent node can send datagrams directly to the mobile node using a routing header. A datagram with a routing header contains the care-of address as the destination address and the home address in the routing extension header as the final destination. Thus the datagram will be normally routed to the care-of address. When the mobile node receives a datagram with a routing header it swaps the final destination with the destination address field. The home address option and the routing header make the mobility transparent with direct routing. [4]

5.2 The Effect on Routing

By using direct routes in both directions the consumption of network resources is minimized.  The 40-byte IPv6 headers consume extra bandwidth when compared to 20 byte IPv4 headers. However the use of routing header and home address option removes the need for constant tunneling, thus decreasing the bandwidth consumption. Although they both add overhead to packets they still are considerably smaller than  IPv6 headers, which would be used in tunneling. The destination options used for signaling can be piggybacked [4] which decreases the signaling overhead considerably, since the options are relatively small when compared to UDP packets.

5.3 The effect on handoffs

The IPv6 mobility support provides the previous router notification mechanism, with which the amount of lost of packets in handoffs can be reduced [4]. In IPv6 the mobile node sends a binding update directly to the previous router, which consumes more bandwidth but is faster than the mechanism used with Mobile IPv4 route optimization.

5.4 Problems solved

Mobile IPv6 provides improvements on routing and signaling efficiency. As the signaling can be mostly piggybacked on data packets there will be considerably less signaling overhead between the mobile node and the correspondent nodes than in mobile IPv4 route optimization between the home agent and the correspondent nodes. The minimum requirements for the correspondent node provide at least triangle routing even in the worst case, since care-of address can be used as the source address. Hosts that are likely to communicate with mobile nodes will probably implement the binding cache and communicate directly with the mobile node. In both cases the routing saves network capacity and decreases delays, when compared to reverse bi-directional tunneling between the mobile node and correspondent node.

The key management problem is not solved Mobile IPv6 does not solve the key management problem, but the integration of IPSec [5] into IPv6 is likely to result in support for key management protocols in most operating systems implementing IPv6.

6 Conclusion

With the increasing number of mobile hosts optimal routes are a goal worth striving for. Route optimization provides means for direct routes between the mobile node and its correspondent nodes. Technology-wise it provides a good framework of techniques to support direct routes. Deployment-wise it is rather problematic. It requires rather large changes to the operating systems of the correspondent nodes. It also requires a trust relationship between the correspondent node and the home agent of the mobile node. As a result of these requirements it probably will not be widely deployed in the near future. The situation will most likely change with the possible transition to IPv6, since mobility support will be a part of the protocol specification at that time. Thus route optimization will probably gain widespread support only via Mobile IPv6.

Glossary

CN Correspondent node, any node communicating with the mobile node
FA Foreign agent provides mobility services to mobile nodes in a foreign network
HA Home agent provides mobility services in the mobile node's home network
IP Internet protocol is the network layer protocol used in the TCP/IP protocol suite.
IPSec IP layer security protocol
ISAKMP Internet security association and key management protocol is used to establish security associations
LAN Local area network is a high speed physical network connecting nodes
MN Mobile node is a node capable of changing its location invisibly to any transport level connections
SKIP Simple Key Management for Internet Protocols
TCP Transport control protocol is a connection oriented transport protocol
UDP Universal datagram protocol, a connectionless transport protocol
WAN Wide area network is a relatively low speed physical network

References

[1] Degermark M., Nordgren B., Pink S., IP Header Compression RFC 2507, February 1999
<http://www.ietf.org/rfc/rfc2507.txt>
[2] Droms R., Dynamic Host Configuration Protocol, RFC 2131, March 1997 
< http://www.ietf.org/rfc/rfc2131.txt >
[3] Ferguson P., Senie D., Network Ingress Filtering: Defeating Denial of Service Attacks which employ IP Source Address Spoofing, RFC 2267, January 1998 
<http://www.ietf.org/rfc/rfc2267.txt>
[4] Johnson D., Perkins C., Mobility Support in IPv6, work in progress, 22 October 1999 
<http://www.ietf.org/internet-drafts/draft-ietf-mobileip-ipv6-09.txt>
[5] Kent S., Atkinson R., Security Architecture for the Internet Protocol, RFC 2401, November 1999 
<http://www.ietf.org/rfc/rfc2401.txt>
[6] Maughan D., Schertler M., Schneider M., Turner J., Internet Security Association and Key
Management Protocol (ISAKMP), RFC 2408, November 1998
<http://www.ietf.org/rfc/rfc2408.txt>
[7] Montenegro G., Gupta V., Sun's SKIP Firewall Traversal for Mobile IP, RFC 2356, June 1998
<http://www.ietf.org/rfc/rfc2356.txt>
[8] Montenegro G., Reverse Tunneling for Mobile IP, RFC 2344, May 1998 
<http://www.ietf.org/rfc/rfc2344.txt>
[9] Perkins C., IP Mobility Support for IPv4, RFC 2002, October 1996 
<http://www.ietf.org/internet-drafts/draft-ietf-mobileip-rfc2002-bis-00.txt >
[10] Perkins C., Minimal Encapsulation within IP, RFC 2004, October 1996 
<http://www.ietf.org/rfc/rfc2004.txt>
[11]  Perkins C., Johnson D., Route Optimization in Mobile IP, work in progress 
< http://www.ietf.org/internet-drafts/draft-ietf-mobileip-optim-09.txt >
[12] Van Jacobson, Congestion Avoidance and Control.  In Proceedings of the SIGCOMM '88 Symposium:  Communications Architectures & Protocols, pages 314--329, August 1988.

Further Information

  1. A good introductory paper on mobility issues and Mobile IP by Charles Perkins:

    2. < http://www.baltzer.nl/monet/articlesfree/1998/3-4/mnt071.pdf
  2. Comer's book, Internetworking with TCP/IP, published by Prentice Hall, provides a good introduction to the TCP/IP protocol suite.
  3. The current IETF-drafts and RFCs can be found from IETF's web site:

    4. < http://www.ietf.org/html.charters/mobileip-charter.html >
  4. More in-depth information can be found from the mobile IP mail archive at Nortel networks web site:

    5. < http://www17.nortelnetworks.com/archives/mobile-ip.html >