ServerCast: efficient cooperative bulk data distribution scheme for content distribution networks

We study bulk data distribution schemes for content distribution networks (CDNs) using application-level overlay. Given the outbound bandwidth of all the CDN edge servers, a fluid-flow based analytical model is constructed to derive the optimal bandwidth allocations at the origin server and among all the edge servers. From which, a lower bound on the content update time is obtained. In order to have an efficient and practical scheme for bulk data distribution, ServerCast is designed to take advantages of the analytical results. Simulations are conducted to evaluate the performance of ServerCast. We show that ServerCast not only outperforms the two existing schemes, fastreplica and multiple unicast, but also yields a content update time within 18% of the derived lower bound.