Transient Analysis of Queues for Peer-based Multimedia Content Delivery

Consider a firm that sells on-line multimedia content. In order to manage costs and quality of service, this firm maintains a peer-network whereby users download files from other peers that have downloaded them in the past. The scenario can be modeled as a queueing system where the number of servers varies over time. We develop analytical models based on fluid and diffusion approximations to analyze transient system performance. Using the same approxi-mations, we also analyze the steady-state behavior of such a network. We, however, find that existing fluid and diffusion approximations are inaccurate for transient analysis. To address this shortcoming, we provide a novel Gaussian-based adjustment and significantly improve accuracy in approximations. Further, models used in this research can be extended seamlessly to the case where system parameters (e.g. arrival rates and service rates) are time-varying. We provide several numerical examples to show how our adjusted models work for transient analysis.