The objective of this research is to carry out a novel study of network survivability across layers to deal with cascading failures in layered networks. This research will be in the context of IP-over-WDM (wavelength division multiplexing) optical networks, with a focus on multiple failures in the physical (optical) layer and their consequences at the higher layer, namely the IP layer. Specifically, the broad scope of the project covers i) Survivable logical topology mapping under multiple failures, ii) Logical topology mapping for guaranteed survivability, iii) Logical topology mapping under multiple constraints, and iv) A generalized theory of flows across layers, capacity of survivable logical topologies and related algorithmic challenges.
The intellectual merits of the proposed research lie in developing unifying theories and methodologies that will make significant advances to the understanding of cross-layer survivability issues, and providing the theoretical foundation for future advances in the general area of cross-layer design and optimization. The research team will build these theories on modern advances in graph theory, mathematical programming and algorithm design. Innovative algorithmic techniques based on advanced data structures and computer algorithms such as approximation techniques will be developed. A generalized theory of cross layer flows that will go well beyond the widely used classical theory of single layer flows will be developed.
Broader Impact: WDM Optical networks form the critical backbone of all modern communication networks and systems. Modern communication systems consist of multiple physical implementations communicating via layered protocols. As such, a single failure at one layer may lead to cascading failures, i.e., failures at the physical layer lead to failures at the logical layer. Research in this area of cross-layer survivability is still in its infancy. Though IP-Over-WDM networks will provide the context for this research, the theory of cross layer flows and the algorithmic (in particular approximation) techniques that will be developed will have multidisciplinary value spanning several STEM related areas: computer science, electrical engineering, graph theory, mathematical programming, and the emerging area of network science. Besides extending the frontiers of knowledge in cross layer theory, the research will have significant educational value in training highly skilled researchers for research and development in cutting edge technologies in different areas of information technology. The training program will have a specific focus on training undergraduate students for future leadership in STEM areas.