Abstract
A wireless communications network maintains per-user information about each subscriber in user profiles. A user profile contains information such as current location of the user, authentication information, billing information, and other advanced feature information. User profiles are stored in databases connected to the signaling network. When a person makes a call, the network must perform, for call delivery purposes, a location lookup to obtain the callee's current location from a copy of the user profile. When a user moves from one region to another, the network must modify information in the user profile to track the person's new location. In this research, we address the location management problem---the efficient retrieval and maintenance of location information in user profiles---for large wireless communications network.

Current location management standards in cellular telephony, IS-41 and GSM, are limited by geographical phone numbers, i.e. users cannot maintain the same phone numbers if they have relocated and changed service providers. A lifelong permanent phone number is a highly attractive feature that has been proposed for future wireless communications services. Any proposed technique for supporting lifelong numbering must be scalable and efficient as measured by location lookup latency and the loads presented to profile databases and the signaling network.

In this talk, we will present a new location management technique, Hierarchical Online Parametric ProfilE Replication (HOPPER), that was developed for supporting lifelong numbering in a scalable and efficient manner. We achieve lifelong numbering and scalability with an hierarchical organization of databases. By using a selective replication technique, we obtain better location lookup performance than other proposals while incuring fewer database transactions and lower network signaling loads.

Research into a modeling framework for realistic performance evaluation of location management schemes in wireless networks is severely lacking. Since user calling and mobility patterns are not well understood, performance studies in the literature have used simple models that incompletely characterize user behavior. Realistic user behavior models are crucial to performance evaluation and for establishing the relative advantages of proposed location management techniques. Through statistical analysis of actual call traffic traces, vehicle and airplane traffic data, and government transportation surveys, we have developed a realistic teletraffic modeling framework for large wireless communications networks. Our framework incorporates callee distributions, different call traffic types, and common user mobility patterns.

Using large scale network simulations, we have studied the performance of our proposal and other techniques using a model for the ten-largest cities in the United States and based on our teletraffic research. We will show simulation results comparing profile lookup latency, database access requirements, and signaling network loads.