Information Geometric Network Architecture for Heterogeneous Network Management

We present a unifying architecture for modeling, optimization and design for heterogeneous networks that can effectively address the existing limitations inherited from the generalized network utility maximization (GNUM) framework. In general, GNUM framework is a unifying mathematical model as it combines different objectives and constraints from different layers into a single global optimization formulation. The solution to this optimization problem automatically provides the benchmark for all layering schemes and hence is the ultimate upper bound on the overall performance of the network. However, GNUM lacks vital components to effectively formulate the realities observed in heterogeneous networks. In this work, we address the key issue of nonconvexity observed in many GNUM formulations. We propose a new perspective to analyze and discover decomposition structures that the GNUM problems exhibit. Theory of Information Geometry and in particular Differential Geometry is central to our approach as we make use of the geometric properties of the problem that can lead to valuable insights and lessons regarding the general structure of the GNUM problems. It is believed that the proposed architecture by innovatively incorporating the theory of Information Geometry into to GNUM formulation potentially provides the necessary tool currently missing for effective management of heterogeneous networks. BENEFIT: The proposed information geometric network architecture and various optimization mechanisms provide a solid solution that can facilitate systematic investigation of holistic network modeling, analysis and design. We expect to produce a comprehensive optimization and simulation toolbox equipped with our flexible framework and efficient algorithms for military as well as researchers in the networking community to investigate homogeneous and optimized network management problems. The models and the software tool can be applied to a broad range of military networks including war-time command and control, real-time surveillance network, homeland security, etc. Other potential commercial applications include border and coast patrol, law enforcement agency, emergency control center and various civil applications. In essence, the product resulting from this effort will be applicable to virtually all networks. The market is quite large and still developing due to the fact that there is no systematic network modeling and design tool that can render the much needed “holistic” investigation of network modeling and design. The aggregated commercial market size can be much larger than that of military applications. IAI is more than a “think tank”, and we have actively pursued with our partners the application of our technologies into actual products in the past. For this proposed effort, in particular, we strongly believe that our work provides the solution needed in practice, and we will invest significant amount of efforts and internal funding to develop the software tool in the long-run. It is also reasonable to expect a source of revenue from service contracts related with the actual development of such product of systematic network modeling and design tools. In addition, IAI will closely work with our partners and collaborator companies such as Lockheed, Boeing, BAE systems, Raytheon, and Telcordia to transfer this technology into the military and commercial world.