Air-Caching: Adaptive Hybrid Data Delivery

Konstantinos Stathatos

Doctoral Dissertation, Number: CSHCN PhD 99-1, Year: 1998, Advisors: Nick Roussopoulos and John S. Baras

Abstract

With the immense popularity of the Web, the world is witnessing an unprecedented demand for on-line data services. A growing number of applications require timely data delivery from information producers to thousands of information consumers. At the same time, the Internet is evolving towards an information superhighway that incorporates a wide mixture of existing and emerging communication technologies, including wireless, mobile, and hybrid networking. For this new computing landscape, this thesis advocates creating highly scalable data services based on adaptive hybrid data delivery. It introduces air-caching, a technique that eectively integrates broadcasting for massive dissemination of popular data, and unicasting for upon-request delivery of the rest. It describes the special properties, performance goals, and challenges of air-caching. Then, it presents adaptive cache management techniques for three dierent settings: servicing large numbers of data requests over a heavily accessed databases, propagating 1 data updates to mobile clients intermittently connected to information sources, and implementing publish/subscribe services again in the context of mobile computing. In all cases, performance experiments demonstrate the scalability, eciency, and versatility of this technique, even under rapidly changing data access patterns.