"Resource Scheduling for Composite Multimedia Objects"

by Minos N. Garofalakis, Yannis E. Ioannidis, and Banu Özden.
Proceedings of VLDB'98, New York City, USA, August 1998, pp. 74-85.



Abstract

Scheduling algorithms for composite multimedia presentations need to ensure that the user-defined synchronization constraints for the various presentation components are met. This requirement gives rise to task models that are significantly more complex than the models employed in scheduling theory and practice. In this paper, we formulate the resource scheduling problems for composite multimedia objects and develop novel efficient scheduling algorithms drawing on a number of techniques from pattern matching and multiprocessor scheduling. Our formulation is based on a novel sequence packing problem, where the goal is to superimpose numeric sequences (representing the objects' resource needs as a function of time) within a fixed capacity bin (representing the server's resource capacity). Given the intractability of the problem, we propose heuristic solutions using a two-step approach. First, we present a "basic step" method for packing two composite object sequences into a single, combined sequence. Second, we show how this basic step can be employed within different scheduling algorithms to obtain a playout schedule for multiple objects. More specifically, we present an algorithm based on Graham's list-scheduling method that is provably near-optimal for monotonic object sequences. We also suggest a number of optimizations on the base list-scheduling scheme. Preliminary experimental results confirm the effectiveness of our approach.

[ camera-ready paper (pdf) (ps.gz) | extended version (ps.gz) | my talk slides (ps.gz) ]


Copyright © 1998, VLDB Endowment. Permission to copy without fee all or part of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the VLDB copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the Very Large Data Base Endowment. To copy otherwise, or to republish, requires a fee and/or special permission from the Endowment.