University of Houston
Department of Computer Science


In partial fulfillment of the Requirements for the Degree of
Master of Science


Lei Yu
will defend his thesis

  A Restrained Hierarchical Stream Merging Protocol for Video on Demand



Abstract

 Video-on-demand (VOD) services provide a much wider services at a higher level of convenience than any of its predecessors, like cable television, video rentals and pay-per-view. But VOD has not succeeded as predicated. The main reason for this lack of the success is the high cost of the service.  Video is a popular but expensive media. The main reasons are its size and sensitivity of delay and loss.

Hierarchical multicast stream merging (HMSM) is one of those proposals focusing on lowering the cost of VOD server bandwidth requirements significantly. It provides instant access to the videos and does not require the customer set top box (STB) to receive video data on more than two channels. 

We have presented a new stream merging policy aiming at reducing the bandwidth requirements of the HMSM protocol.  Our retrained earliest reachable merge target (RERMT) policy improves upon the earliest reachable merge target (ERMT) policy by redefining complete stream when the new client arrives, according to the time interval between the incoming stream and the first available stream.   If this time interval is smaller than the sum of the interval between last available stream and the original complete stream and the time interval between the first available stream and the last stream, the first available stream becomes the new complete stream.  Otherwise, the incoming stream is the new complete stream. Experimental results have indicated that RERMT outperforms all existing HMSM merging policies at all request arrival rate. At arrival rates more than 20 requests per hour, it can improve the average server bandwidth and server bandwidth by up to 38 percent than ERMT.

 

Date: Monday, November 24, 2003
Time: 1:30 PM
Place: 550-PGH


Faculty, students, and the general public are invited.
Thesis Advisor: Dr. Jehan-François Pâris