Adaptive Data Broadcast in Hybrid Networks

K. Stathatos, N. Roussopoulos, J.S. Baras

Twenty–third International Conference on Very Large Databases, August 1997, Athens, Greece.

Abstract

With the immense popularity of the Web, the world is witnessing an unprecedented demand for data services. At the same time, the Internet is evolving towards an information super-highway that incorporates a wide mixture of existing and emerging communication technologies, including wireless, mobile, and hybrid networking. Taking advantage of these new technologies, we are proposing a hybrid scheme which effectively combines broadcast for massive data dissemination and unicast for individual data delivery. In this paper, we describe a technique that uses the broadcast medium for storage of frequently requested data, and an algorithm that continuously adapts the broadcast content to match the hot-spot of the database. We show that the hot-spot can be accurately obtained by the monitoring the "broadcast misses" observed through direct requests. This is a departure from other broadcast-based systems which rely on efficient scheduling based on precompiled user profiles. We also show that the proposed scheme performs effectively even under very dynamic and rapidly changing workloads. Extensive simulation results demonstrate both the scalabillity and versatility of the technique.