Replication Algorithms for Scalable Streaming Media in Content Delivery Networks

1. Introduction CDN (Content Delivery Networks) improves end-user performance by replicating web contents on a group of geographically distributed servers. However, current replica strategies in CDN are to simply and repeatedly keep the complete replica of the original object on many content servers. Disadvantages of this method are as follows: to repeatedly store the same large-sized object into different content servers consumes too much server space. Also because some of the content servers are not always requested by clients, to keep replicas on these servers waste storage cost. It is more serious for replicating some large-sized objects such as streaming media, e.g. high quality video, which are being distributed over the Internet more and more. There have been many other published studies on streaming media. However, almost all researches are about how to replace or cache streaming in a single proxy cache. How to efficiently replicate and distribute streaming media in a group of servers such as CDN has not been mentioned. Recently researchers found that scalable (layered) streaming media is appropriate for the Internet because of its better flexibility and functionality. In this paper, we therefore discuss about a scheme to replicate scalable streaming contents in CDN. Our work focuses on how to replicate different layers of different streams into different content servers to save the system resources and improve the user delay at the same time. We firstly carry out a theoretical analysis of storage cost, capacity limit and access distribution of scalable streaming contents. Based on these analyses we then propose a replication algorithm in which not only the popularities of streams and servers but also the current network topology is considered. Through simulations, we check the performance of our proposal when the related parameters are changed. Simulation results show that our proposal can efficiently minimize user delay and network traffic. 2. Scalable Streaming Scalable (or layered) streaming is supported by some standards, such as MPEG-2, MPEG-4 and H.264. Basically, the original signal is coded into several layers, from the lowest (base) to the highest (enhancement) layers. There are some advantages by using scalable streaming. 1) A sender can easily control compression ratio by dropping some of layers according to network condition. 2) A receiver can select the number of layers according to its own network condition, known as receiver-driven layered multicast. 3) A content server can keep only some of layers instead of the whole original …