A realist's perspective on modeling, analysis and design of wide area networks: recent work and emerging issues

Building software-based tools for the analysis and design of large-scale, wide area communication networks requires more than a judicial application of graph theory, optimization and queueing; it requires an effective "systems engineered" approach to managing real-world constraints and change throughout the life-cycle of competing networking technologies. This talk will examine the role of theory, realism, and process in roughly ten years work by the author and colleagues in developing tools for modeling, analysis, design of wide area networks. Starting with the philosophical and methodological underpinnings of the work, the talk will: ~ describe a range of real-world network analysis and design scenarios that arise in the life cycles of these networks, ~ characterize various design problems associated with these scenarios, ~ trace the development and continued evolution of a collection of tools for addressing the design problems, and ~ illustrate some new scenarios (e.g., ATM) driving current research. Of particular emphasis throughout this work is a philosophical bent towards realistic problem formulations and solution techniques which rapidly yield implementable results for large backbone networks employing distributed, adaptive routing. A principal product of the this research is a large scale software system known as M11DAS (Mod ular Information Infrastructure Design and Analysis System). M11DAS (and its UVa-developed ancestor, DESI) is used by the Center for Engineering, Defense Information Systems Agency as its principal data backbone design and analysis tool and is one of the key tools used in supporting planning for DISA's Defense Information System Network architecture. Recently, the tool has also been adapted under FAA sponsorship through MITRE's Center for Advanced Aviation System Development (CAASD) to support analysis and design of the FAA's NADIN 11 network.