Large-scale optimization for data placement problem

Large-scale optimization of combinatorial problems is one of the most challenging areas. These problems are characterized by large sets of data (variables and constraints). In this thesis, we study large-scale optimization of the data placement problem with zero storage cost. The goal in the data placement problem is to find the placement of data objects in a set of fixed capacity caches in a network to optimize the latency of access. Data placement problem arises naturally in the design of content distribution networks. We report on an empirical study of the upper bound and the lower bound of this problem for large sized instances. We also study a semi-Lagrangean relaxation of a closely related k-median problem. In this thesis, we study the theory and practice of approximation algorithm for the data placement problem and the k-median problem