Heuristics for the Design and Optimization of Streaming Content Distribution Networks

The design of efficient and scalable streaming Content Distribution Networks (CDNs) is an open problem of great economic interest. A key decision is the number of replica servers to use and their location to balance network bandwidth and server costs. This problem is highly complex, since it is interrelated to the streaming media delivery protocols, client redirecting decisions, and the content delivery routing problem. To address this design problem, we develop heuristics for the server placement, the server selection, and the routing problem. In contrast to previous work, we study both periodic broadcast and hierarchical merging protocols using multiple media files with different request rates. The heuristics lead to highly accurate and fast approximations. We also analyze the relationship between the start-up delay caused by some delivery protocols, the bandwidth consumption, and the number of replica servers, providing complete and fully implementable solutions to the design of streaming CDNs.