Mobility management in wireless networks is an important task in order to keep connectivity with roaming users at anytime. Mobile IP1, which is a standard proposed by the Internet Engineering Task Force (IETF), can serve as the global mobility management in the future heterogeneous wireless networks.

Mobile IP uses the home agent (HA) and foreign agent (FA) to maintain the mobility of a mobile node (MN). The HA maintains the address binding of an MN, and the address binding is a mapping between the permanent home address to the care-of address (CoA) temporally borrowed from an FA.

Mobile IP suffers from several problems3-5 such as the triangular routing, frequent and long distant registration updates, and single point of failures.

System Overview
Researchers are proposing multiple solution and one of them is summarized below. Suppose that each subnet is associated with an FA. Several subnets would constitute a domain which is associated with a GFA. The functional overview of the proposed architecture is depicted below.

P2P-based Mobile IP
To avoid the problems of single point of failure and overloaded traffic on the HA, we organize all the existing HAs into a P2P network. With the essential feature of P2P networks, one HA can freely join (when newly installed) and leave (when failed) the network. We would dynamically select an HA for an MN from the P2P network, which is close to the MN. If the selected HA is heavy loaded, we would seek another light loaded one in the neighborhood. With a little modification on the existing P2P lookup mechanism, we can efficiently locate the HA that is selected for a particular MN in the P2P network.

When an MN moves within the same domain, the registration update is locally performed to the GFA. Only whenever the MN moves to another domain, the registration update to the HA is performed. Meanwhile, we may select a new HA that is close to the MN for reducing the long distant registration update.

When a CN (Corresponding Node) would like to connect to an MN, it issues a query to DNS where the P2P lookup mechanism is triggered to locate the MN’s HA. The found HA would return the location of GFA the MN is currently located in to the CN. As a result, the CN can directly establish a connection to the GFA and this connection would be further redirected to the MN.

We claim that this architecture can have the following advantages:

Update locality. The frequent registration updates due to the MN’s movement of small scope will be partially localized by the regional registration technique. Moreover, the periodical registration updates to the HA during the binding renewal period would be cost saving, because we have selected a near HA to the MN.

Load balancing. The DNS does not perform the complex name resolution for an MN. Instead, the DNS only provides the entry point to the P2P network and triggers the P2P lookup mechanism to find the MN’s HA. We put the burden of the complex name resolution on the P2P network where the actual execution would be distributed to the nodes involved in the P2P network. The set of HAs in the P2P network will work together and can migrate the workload with each other.

Self-administration. Each system operator can administer its own P2P network, which facilities the prevention of binding data from the revelation to other system operators. Also, a system operator can freely increase or decrease the number of HAs depending on the amount of users that are served.

For more details, please download the paper from:

http://www.csie.ndhu.edu.tw/~robert/paper/MWCN2004.pdf