An Optimization Model to Determine Data Center Locations for the Army Enterprise

The Army within the Continental U.S. (CONUS) consists of many installations having a diverse range of user populations and computing hardware/software. Currently, most installations process their own common-user information technology services. As part of a move towards net-centric operations, the Army is looking to consolidate servers and applications at a few data centers known as Area Processing Centers (APCs). The benefits will include improved Warfighters´ operational efficiency, reduced Total Cost of Ownership, and enhanced information security posture of the Army enterprise. To move forward, the Army must determine the optimal number and locations of its APCs as well as optimal dual-homing arrangements for all installations. Due to lack of reliable prediction of the impact to the backbone capacity and application latency, we try to solve a surrogate problem based on notions of equity in demand load and distance traveled between clients and servers. The suggested location and homing solutions may not be optimal in terms of future capacity usage and user experience, but the approach provides a quick and practical method to make location recommendations. An optimization model was developed to recommend APC locations and installation homing arrangements such that the total demand-weighted distance traveled is minimized and the load on any APC can be as small as possible. This paper presents an integer programming model for the problem and demonstrates how to use the model in a practical, iterative process to make facility location and demand homing decisions for any general setting.