Joint optimization of flow allocation and data center placement in multi-service networks

摘要

We present a framework for jointly optimizing the data center placement and network flow allocation in multi-service networks. The optimization is derived within a cost-performance formulation that aims at minimizing the overall operating cost of the network over which the services are provided, given a set of end-user performance requirements. The problem formulation represents a linear programming that is efficiently solved via a convex optimization technique that we design to this end. The generality of the formulation allows its application to other important problems, such as green networking and transportation of hazardous materials. We introduce the concept of cost performance versus service requirements lower-convex hull that can be employed to study the operational efficiency of various data center location and network flow allocation configuration points. We investigate the performance of the optimization framework via simulation experiments and compare its efficiency against two heuristic reference schemes. We demonstrate substantial operational cost savings or conversely service performance advances, for the same overall operating cost.