|
|
|
Mary K. Vernon Department of Computer Science University of Wisconsin-Madison |
|
This talk will describe the SWORD project, which is developing next generation data delivery techniques for the Web and other (satellite/cable) on-demand data delivery systems that contain large, very popular, widely-shared data files. Such files might include news clips, popular product advertisements, television shows, medical and recreational information services, distance education content, and perhaps even full-length movies.
The talk will include a discussion of current delivery techniques, our recent research results that provide improved techniques, and some of the open research issues that we're currently addressing. The improved techniques that will be discussed include (1) methods for segmenting the data and scheduling segment multicasts to improve client cost-sharing, and (2) methods for determining the optimized set of initial segments and complete files to cache at the regional/proxy servers. Together these techniques greatly reduce the server and network bandwidth required for quick response and continuous delivery to clients. We are currently implementing a prototype multimedia server that allows us to address some of the practical implementation issues for next generation multimedia servers that might employ the new delivery techniques.
This work is joint with Prof. Derek Eager (University of Saskatchewan). The optimized caching work is also joint with Prof. Michael Ferris (Univ. of Wisconsin). Two papers that the talk is based on are linked to the SWORD web page, http://www.cs.wisc.edu/~vernon/sword.html