Caching Broadcasted Data for Soft Real-Time Transactions

摘要

Data broadcast is an efficient communication model when a large number of clients request data from a server in an Internet or wireless computing environment. While scheduling algorithms are studied to reduce clients' average waiting time for requested data to pass by on the air. caching strategies are also designed to memorize some data broadcasted earlier locally at each client to allow later requests to these data to be satisfied immediately. Existing caching strategies, which are proved to efficiently reduce the average waiting time, are designed to improve the performance of single data operations. In this paper, we introduce caching schemes that favor soft real-time multi-operation transactions. Instead of overall average response time of single operations used in existing caching methods, the performance measurement of the caching policies studied in this paper is the percentage of real-time client transactions that meet deadlines. Performance evaluation result shows our new strategies have many advantages and provide better real-time performance than existing methods.