Scaling Peer-to-Peer Video-on-Demand systems using helpers

Abstract

The throughput of Peer-to-Peer (P2P) Video-on-Demand (VoD) systems is typically capped by the users' aggregate upload bandwidth [1]. The drastic increase in the popularity of VoD and the demand of higher quality content has thus placed substantial burden on the content servers. We investigate a novel P2P VoD architecture that leverages idle Internet resources, which we call helpers, to provide a scalable solution to P2P VoD systems. Helpers are volatile in nature, and can be individually unreliable. However, we investigate the statistical aggregation of a large number of helpers to guarantee quality of service. Since helpers do not come with ¿free¿ preloaded content, trade-offs between how much a helper should download and how much it can aid the system need to be explored. In this paper, the optimal steady-state design parameters are derived to maximize the helpers' upload bandwidth utilization. Packet level simulations have verified the efficiency of the system. In a typical scenario of 240 users and a required theoretical minimum of 120 helpers with an average upload bandwidth of 256 kbps, a streaming rate of 384 kbps can be sustained with < 2% relative server load. Results also show that the system is robust to helper churn.