Effect of failures on optimal location management algorithms

The current IS-41 standard PCS architecture uses a centralized database, the Home Location Register (HLR), to store service and location information of each mobile registered in the PCS network. However this is a single point of failure, therefore distributed database architectures have been proposed as possible architectures for future mobile networks. The performance of such architectures improves when the load on the databases is balanced. Though distributed architectures offer increased fault tolerance, they suffer from increased query delay (number of databases to be queried to retrieve the required information). The query data further increases in the presence of database failures. In this paper we derive bound for the performance of load balanced location management algorithms in the presence of database failures for three standard performance metrics, worst-case query, delay average query delay, and call blocking probability. A load balanced algorithm tolerating database failures is presented, and its performance measured using the same performance metrics. A robust parallel location management algorithm which queries databases in parallel is then presented, and we derive expressions for the average query delay for the algorithm in presence of failures.