If the Internet is to fulfill its promise of being the next great communication medium, we need to build content delivery networks (CDNs, for short) that deliver web and streaming content to end-users with high performance, infinite scalability, and low cost. At the very heart of a CDN is a core set of algorithms that sense the network and the internet as whole, efficiently routes content from its source to edge servers, and finally serves the content to end-users from their respective optimal edge servers. This project builds the scientific foundation necessary to meet the significant challenges of designing a next-generation CDN. Specifically, the project expects to make the following major research contributions: (1) accurate models for costs and a precise characterization of the cost/performance tradeoffs in CDN design, (2) novel algorithms for core CDN functionalities that are provably high performing, scalable, and cost-effective, (3) empirical validation of the proposed algorithms using extensive data sets collected from hundreds of vantage points on the internet, and (4) technology transfer of some of the fruits of the research to industry. In addition to research, the project is expected make the following major educational contributions: (1) course offering on building large-scale distributed systems (such as CDNs) that bring together and cross-pollinate ideas from both academic research and commercial system building, and (2) mentoring and training of graduate and undergraduate students in cutting-edge distributed system technologies.
