Some Distributed Approaches to the Service Facility Location Problem in Dynamic and Complex Networks

The need to efficiently accommodate over the Internet the ever exploding (user-generated) content and services, calls for the development of service placement schemes that are distributed and of low complexity. As the derivation of the optimal placement in such environments is prohibitive due to the global topology and demand requirement and the large scale and dynamicity of the environment, feasible and efficient solutions of low complexity are necessary even at the expense of non-guaranteed optimality. This chapter presents three such approaches that migrate the service along cost-reducing paths by utilizing topology and demand information that is strictly local or confined to a small neighborhood: the neighbor hopping migration requires strictly local information and guarantees optimality for topologies of unique shortest path tree; the r-hop neighborhood migration appears to be more effective for general topologies and can also address jointly the derivation of both the number and locations of services to be deployed; the generalized neighborhood migration approach opens up new possibilities in defining localities, other than topological ones, that contain the most relevant candidates for the optimal placement, by exploiting emerging metrics and structures associated with complex and social networks. The underlying assumptions, strengths, efficiency and applicability of each of these approaches are discussed and some indicative results are shown.