Customized Newspaper Broadcast: Data Broadcast with Dependencies

Broadcasting has been proved to be an efficient means of disseminating data in wireless communication environments (such as Satellite, mobile phone networks; other typical broadcast networks are Videotext systems). Recent works provide strong evidence that correlationbased broadcast can significantly improve the average service time of broadcast systems. Most of the research on data broadcasting was done under the assumption that user requests are for a single item at a time and are independent of each other. However in many real world applications, such as web servers, dependencies exist among the data items, for instance: web pages on a server usually share a lot of items such as logos, style sheets, title-bar... and all these components have to be downloaded together when any individual page is requested. Such web server could take advantage of the correlations between the components of the pages, to speed up the broadcast of popular web pages. This paper presents a theoretical analysis of data dependencies and provides a polynomial time 4-approximation as well as theoretical proofs that our correlationbased approach can improve by an arbitrary factor the performances of the system. To our knowledge, our solutions are the first provably efficient algorithms to deal with dependencies involving more than two data items. Topics: Approximation algorithms, Wireless and Push-based broadcast scheduling, Bluetooth and Satellite networks.