A framework for synchronized delivery of documents in a multimedia Web environment

Zafar Ali, Purdue University

Abstract

The traffic patterns observed in a multimedia Web environment are quite different from the traffic variations experienced in communicating a single media stream. In particular, traffic variations within a single multimedia object are mainly due to inter frame compression. However, traffic generated by a multimedia server has varying capacity requirement at browsing, document and object levels. ^ The browsing level variations in the network capacity usage are caused by the user's interactions. In this dissertation, neuro-fuzzy logic is used to model the workload generated in the Web browsing environment. A set of parameters that are used to characterize the work-load generated by the multimedia server is proposed. Consequently, a Neuro-Fuzzy Scheduler (NFS) that makes an intelligent compromise among multi-criteria by properly combining some elementary scheduling heuristics, is presented. Performance of the NFS is compared with several known heuristics and a branch and bound algorithm. The results show that the proposed neuro-fuzzy scheduler can dynamically adjust to the varying work-load quite well. ^ The document level variations in the network capacity usage are due to the changing level of concurrency and the characteristics of the component objects within a multimedia document (MMD). A scheme to quantify the amortized capacity and destination buffering requirement of an MMD is presented. Trade-offs between network resources and user's quality of acceptance are also studied. ^ The characterization of MMDs is based on the assumption that the object level variation for the MMD can be modeled using the effective capacity approximation. An algorithm for modeling traffic generated by a single multimedia object, based on the effective capacity approximation, is proposed. The scalability aspects of the proposed algorithm are compared with some of the previously known schemes for effective capacity computation. The performance of the proposed CAC scheme is evaluated using known performance bounds and simulation results. The results show that the scheme scales well with increasing amount of resources (link capacity and buffer size) and accommodates intelligently the mix of traffic offered by sources of diversed burstiness characteristics. ^

Degree

Ph.D.

Advisors

Major Professor: Arif Ghafoor, Purdue University.

Subject Area

Engineering, Electronics and Electrical|Computer Science

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS