Improved Smallest Load First Storage Scheduling Policy for VOD Server Clusters
ZHOU Hang,JI Hai-bo,LI Jun
DOI: https://doi.org/10.3969/j.issn.1000-1220.2007.12.008
2007-01-01
Abstract:In clustered VOD(Video-On-Demand) server systems,storage scheduling policy greatly influences system's request scheduling,load balancing and scalability.However,each invocation of the existing smallest load first(SLF) replica placement algorithm is independent of the previous one,which could result in high cost of reorganization and make SLF impractical in many environments.To resolve this problem,an improved replicas placement algorithm is proposed.By making full use of previous stored replicas,the algorithm reduces the cost of adapting to changes in client demand or server configuration.Experiment shows that using the storage scheduling policy with proposed replica placement algorithm,our system achieves high user request acceptance rate as well as low load imbalance and low storage scheduling cost.