Fault-tolerant analysis for multiple servers movie retrieval strategy for distributed multimedia applications

  • Authors:

  • Bharadwaj Veeravalli;Chaoyang Chen;Viktor K. Prasanna

  • Affiliations:

  • Open Source Software Laboratory, Department of Electrical and Computer Engineering, The National University of Singapore, Singapore, Singapore 117576;Open Source Software Laboratory, Department of Electrical and Computer Engineering, The National University of Singapore, Singapore, Singapore 117576;Department of Electrical Engineering - Systems, EEB-200C, University of Southern California, Los Angeles, USA 90089-2562

  • Venue:
  • Multimedia Tools and Applications
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we address the problem of retrieving a movie from a set of multimedia(MM) servers by the clients on a network. We consider a strategy in which multiple MM servers are deployed by the service provider (SP) to retrieve a requested MM movie to the clients, for minimizing the access time (the waiting time of the client before initiating the playback) and maximizes the system reliability. We design a movie retrieval strategy that explicitly considers issues such as reliability and/or availability factors of the multimedia servers and the communication channels in the problem formulation. We develop a mathematical model for this retrieval strategy and derive an optimal size of each movie portion that is expected to be rendered by each server. We then derive a closed-form expression for the access time of the MM document and the system reliability which gives a trade-off relationship between access time and reliability (availability) of the service by our strategy. We extend our study to investigate on the effect of sequencing of the servers, the order in which movie portions are to be retrieved, to minimize the access time and to maximize the system reliability. With system reliability factors, we identify an optimal sequence, which maximizes system reliability out of all possible retrieval sequences. We then propose two methods to retrieve any missing movie portions upon a server failure during the retrieval process. In order to measure the quality of service provided by the service provider to its customers, we introduce a QoS parameter that can tune the playback rate to avoid any data underflow or overflow situations. Then, from probabilistic perspective, we obtain an estimate of the failure time of a single server and its resulting missing movie portion caused by this server failure. We conduct rigorous simulation experiments to testify all the theoretical findings reported. Illustrative examples are provided for the ease of understanding.