Mobility Management in PCS Networks
Karen Qing Tian
Dept. Electrical Engineering Department
We take the stance that a Universal Personal Identification (UPI) will become an integral
part of future personal communication systems so as to better serve the ever increasing communication needs across
heterogeneous networks and through a variety of diverse communication devices. Under such context, the relevant problem
to be solved is how to locate a user given his or her UPI.
In traditional wireless communication systems, the network keeps track of a user's location through an up-to-date user profile
stored in various databases. A user profile contains not only user's current location information, but also service information,
such as billing and authentication. The two basic operations in mobility management are location update and location lookup. When
a user moves across the boundaries of registration areas, the network updates the user's location information. When a caller places
a call, the network queries the relevant database(s) to obtain the current location and other service information of the callee.
In IS-41 and GSM standards, a user is identified by a geographical phone number that directly points to the database containing
the user profile. However, geographical numbering ties a user's identification to a particular network and geographical location,
therefore can't support UPI.
Hierarchical mobility management techniques have been proposed to support UPI. In addition to using a hierarchical structure, the
performance of mobility management can be further enhanced by using replicas of user profiles which may be kept at various locations.
Replication techniques make profile information more readily available, thus reducing lookup cost and latency, but to keep these
replicas consistent and fresh, they must be updated whenever the user profile is updated. The principle of replication is to replicate
if the benefit of replication is greater than its overhead.
We propose optimal off-line replication algorithms that minimize the network messaging cost based on network structure,
communication link cost, and user traffic statistics. In contrast to an off-line algorithm, an on-line algorithm does not assume
any user traffic statistics, it decides whether to distribute new replicas, or delete existing replicas after serving each request,
all based on the input sequence seen so far. Our work not only generalizes previously proposed on-line algorithms within a unified
framework, we also effectively reduce the overhead incurred by these algorithms, hence making a family of on-line algorithms feasible
in practice. The performance of our profile replication algorithms is studied via large scale network simulations. Both our off-line
and on-line replication algorithms are optimal and perform better than previously proposed threshold-based algorithms.