Tively. the List Based Scheme Performs Better Than

both the broadcast and the pointer based schemes for Prole I over a wide range of call/mobility ratios. The pointer based scheme performs better than the list based scheme only for a high call/mobility ratio in Prole II. Pointer based scheme incurs a penalty for informing the base stations of the new address for each move. The lower the mobility, the lower the cost of pointers. Thus under high call/mobility ratios, the pointer based scheme incurs a lower cost than under low call/mobility ratios. The improvement in performance (decrease in cost) as a result of partitioning is very signicant compared to that of broadcasting without partitioning. For example, with the above user proles and the chosen partitioning, the improvement in performance for moderate to high call/mobility ratios with the pointer and list based schemes is two to three fold over schemes that just employ broadcasting. Further, broadcasting without partitioning (i.e., broadcasting to all base stations under a location server) would perform even worse. The pointer based scheme is suited to users exhibiting random movements within each partition and when the call/mobility ratio is moderate to high (Prole I). Under these circumstances , if a list based scheme is used, then a call will be expensive as one has to try all equally likely locations. However, the list based scheme is suited to proles exhibiting a more determin-istic movement within small partitions and when the call/mobility ratio is low to moderate (Prole II). At low call/mobility ratios (high mobility), the list based scheme does not incur any cost due to updates and when a user is often found in one or two most likely locations, the cost of the call is also cheaper. Hence the list based scheme outper-forms the pointer and the broadcasting schemes under this condition. Simulations demonstrate that prole information can drastically reduce the number of location updates. In the future, we plan to perform more extensive simulations and also extend our cost model to include the costs of storing prole information. We also plan to simulate a dynamic updating protocol, where the decision whether to update the location or not is based on the recent history of the call to mobility ratio. This is analogous to dynamic replication schemes [25] where the number of replicated copies of data depends upon the read-write pattern on the data item. Figure 6: Performance of Various Strategies 12 consider typical proles …