Scheduling design for distributed video-on-demand servers

Abstract

An online request migration scheme and a dynamic content update strategy for distributed video-on-demand (VoD) systems are proposed in this work to reduce the service failure rate and the server-side network bandwidth consumption. By improving the random early migration (REM) scheme that was designed for a centralized VoD server system, we propose a cost-aware REM (CAREM) scheme to reduce the network bandwidth consumption of the migration in a distributed environment. A dynamic content update strategy is also proposed based on the server-video affinity metric to reconfigure video copies assigned to media servers. Extensive simulations show that CAREM together with the dynamic content update strategy can reduce both the failure rate and the server bandwidth consumption.