Techniques for synchronous retrieval of concurrent multimedia I/O streams
Abstract
I/O has traditionally been a limiting factor in the performance of modern high-performance computer systems consisting of CPUs, memories, secondary storage devices, user interfaces, and the network. This I/O bottleneck becomes even more pronounced for currently evolving systems handling digital multimedia, such as audio and video data. In this dissertation, we study the characteristics of multimedia data, and based on these characteristics, we study and develop efficient techniques for the storage and synchronous retrieval of such data. We propose methods to quantify user perceived quality via QOP (quality of presentation) parameters. We combine QOP and OCPN (object composition petri-net) multimedia data modeling techniques to develop techniques to allow efficient synchronous retrieval of multimedia data. Since I/O bandwidth is a precious resource, we develop methods for the allocation of I/O bandwidth among several concurrent multimedia I/O streams. These methods are fair among users and make judicious utilization of the I/O bandwidth. Inter-media and intra-media temporal synchronization at the user-interaction level is an essential concern in multimedia systems. In order to effect such synchronization, we propose several $O(n\ \log\ n + mn)$ (where m is the number of logical I/O channels and n is the number of frames of multimedia data) heuristics to schedule multimedia frames between the disks and destination memories of the multimedia systems. We study the performance of these heuristics via simulations and show the tradeoffs between the system resources, such as memory, and QOP parameters. Finally, we identify promising avenues for future advances in multimedia I/O.
Degree
Ph.D.
Advisors
Ghafoor, Purdue University.
Subject Area
Computer science|Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.