Design of a starvation-free storage system for a multicast video-on-demand system using a double-rate batching policy

Abstract

Current implementations of video-on-demand (VoD) use a dedicated stream to serve an individual customer. This is not cost-effective. In order to increase the system performance, a multicast transmission technique is used in a VoD system. However, the customers would suffer from a long start-up delay in a multicast system. Thus, the double-rate batching policy was proposed to serve the customers as soon as possible. In this policy, new customers are first served by the dedicated streams by doubling the transmission rate so that they can catch up and join into one of the multicast groups. As the transmission rate is changed, the traditional disk scheduling is not suitable for the double-rate batching policy. In this paper, efficient disk scheduling schemes are developed to provide a starvation-free storage system for the multicast VoD system using double-rate batching. The results show that when the request rate is 0.1 arrivals/s, only 22 disks are sufficient to reduce the waiting time to less than 0.5s. In addition, the memory requirement of the scheduling algorithm in the server is also evaluated. It is shown that the proposed schemes require less memory than the traditional VoD system for the popular movies.