Distributing multimedia content to balance quality of service and cost

An increasing number of computer applications today use multimedia content such as images, sound, and video over distributed networks of computers. Of ten, a dispersed set of users, with varying demands, requires ongoing acces s to this content. Effective placement of the multimedia content at differe nt locations/processors thus becomes essential to ensure acceptable quality of service at a reasonable cost. Achieving this requires the consideration of a set of issues quite different from that required for traditional data distribution. These include (a) scale, both in terms of individual objects and in aggregate, (b) importance of form or appearance, making resolution levels an important, controllable variable, and (c) the temporal dimension, placing stringent demands on response time. These concerns make distributi on of multimedia content more than a straightforward extension of tradition al distribution approaches. We develop a model and a supporting approach to facilitate effective distribution of multimedia content, focusing on multi media applications in corporate intranets. The model consists of multiple c riteria to reflect different aspects of quality of service and cost which w e formulate by leveraging variance in resolution levels to capture trade-of fs among these criteria. Since the multiple-criteria allocation model is NP -complete, we propose a decision support approach that generates locally ef ficient solutions using designer-specified targets and evaluates them using fuzzy-set-based heuristics. The complete model and the approach have been implemented in a prototype to ensure feasibility. We demonstrate use of the prototype for a medical imaging application that illustrates applicability and usefulness of our proposals.