| Article ID: | iaor20082092 |
| Country: | United States |
| Volume: | 55 |
| Issue: | 3 |
| Start Page Number: | 518 |
| End Page Number: | 531 |
| Publication Date: | May 2007 |
| Journal: | Operations Research |
| Authors: | Kalvenes Joakim, Keon Neil |
| Keywords: | service, communications, yield management |
This paper considers the provisioning of transmission line bandwidth on a private network with given traffic routing for the purpose of distribution of video-on-demand service with guaranteed end-to-end quality of service. We present an architecture for video-on-demand service delivery and model the assignment of bandwidth in the distribution network as a constrained, nonlinear optimization problem. To solve this optimization problem, we develop three new auction algorithm-based solution procedures. The optimization problem assumes that there is a functional relationship between the maximum acceptable end-to-end delay and the bandwidth requirement for the links in the distribution network. Absent reliable video traffic data models for MPEG-2 format, we sample a large number of DVD-recorded movies to form a basis for randomly generated aggregate traffic streams. The aggregate traffic streams are used in a simulation experiment to measure the maximum transmission buffer occupancy for each given traffic stream for different transmission rates. Based on this simulation experiment, we derive an empirical transmission line provisioning function that guarantees delivery of all video frames without frame loss within a maximum frame delay tolerance. To illustrate the effectiveness of the proposed solution procedure for the bandwidth assignment problem, we solve to near optimality 570 small problem instances under three demand structures and 100 large problem instances with uniformly distributed demand.