Object replication strategies in content distribution networks

Recently, the Internet has witnessed the emergence of content distribution networks (CDNs). In this paper, we study the problem of optimally replicating objects in CDN servers. In our model, each Internet autonomous system (AS) is a node with finite storage capacity for replicating objects. The optimization problem is to replicate objects so that when clients fetch objects from the nearest CDN server with the requested object, the average number of ASs traversed is minimized. We formulate this problem as a combinatorial optimization problem. We show that this optimization problem is NP complete. We develop four natural heuristics and compare them numerically using real Internet topology data. We find that the best results are obtained with heuristics that have all the CDN servers cooperating in making the replication decisions. We also develop a model for studying the benefits of cooperation between nodes, which provides insight into peer-to-peer content distribution.