Survivable Design of Multi-layer Networks Considering Disasters, Failures, Demand Surges and Long Range Dependent Traffic
- Moshe ZUKERMAN (Principal Investigator / Project Coordinator)Department of Electrical Engineering
- Ronald G ADDIE (Co-Investigator)
- Rongping LIN (Co-Investigator)
DescriptionModern societies depend heavily on the Internet and its failure may have grave socioeconomicconsequences, so its resilience is very important. The proposed project aims todevelop novel methodologies for cost-effective design of survivable multi-layercommunications networks to achieve resilience under a wide range of failure scenarios.We will provide near-optimal analytical solutions applicable to multiple transport technologies, and realistically sized networks, where the traffic modeling include constantand variable components. We significantly extend here our earlier work by consideringsurvivability. While this increases complexity and adds scientific challenges, we aim for amethodology that yields accurate, scalable and validated solutions.The design objective is to provide link capacity assignments, with sufficient, but notexcessive, levels of redundancy. We aim for near optimal choices of the routing paths andtransport technologies, so that, the Earnings Before Interest and Taxes (EBIT) is maximized.EBIT is an important financial measure commonly used by businesses for resource allocation.Our cost-metrics include the operational and amortized capital expenditures, the cost of the effect of network resource utilization, and penalties paid when service disruptions occur.These penalties, which allow prioritization of failure/traffic-surge events, were not consideredearlier in survivable multi-layer network optimizations.Efficient heuristic algorithms for routing and capacity assignment will be developedconsidering recovery from failures, or traffic surges. Since cost and routing decisions aremutually dependent, we iterate flow and capacity assignment until convergence is achieved.This will lead to a new approach to the design problem of survivable multi-layered networks.The algorithms will be validated and benchmarked against solutions and bounds obtainedfrom integer linear programming (ILP) formulations. The inclusion of variable (both shortand long rang dependent) traffic streams in a multi-layer network survivability model is anovel (and important) feature that allows many practical design aspects to be considered, e.g.,statistical multiplexing, dynamic connections, and flow-size dependent routing.The strength and novelty of this proposal is in the generality of our problem and the scalability of our solutions that cannot be provided by existing ILP solutions and therefore itrepresents a quantum leap beyond ILP.We will provide a free-access integrated cloud-basedservice platform that will be continuously examined and improved. Considering the dynamicand complex nature of the Internet, outcomes of this project will guide engineers, students,telecommunications professionals and researchers in industry and academia to betterunderstand, plan, manage, upgrade and design survivable networks.
|Effective start/end date||1/08/17 → 20/07/21|