Decentralized approximation algorithm for data placement problem in content delivery networks

Abstract

Recent advancements in Internet technology research, as well as the widespread of commercial content delivery networks, motivates the need for optimization algorithms designed to work in decentralized manner. In this paper we formulate data placement problem, a special case of universal facility location problem with quadratic terms in objective function. The considered combinatorial optimization problem is NP-hard. A randomized algorithm is presented that approximates the solution within factor O(log n) in decentralized environment, assuming asynchronous message passing of bounded sizes. © 2012 IFIP International Federation for Information Processing.