NeTS: Small: Tunable Opto-Electric Network Architectures & Benchmarks

William Marsh Rice University, Houston TX

Investigators

Abstract

Long-term sustainable, energy-efficient, and high-performance cloud computing network infrastructures are crucial to advancing national prosperity and promoting the progress of science. To meet this challenge, this project develops a new family of networks that employs optical circuit switching technologies. The fundamental issues this project seeks to address include the design of the new network structures that are highly resilient and flexible to application requirements, the design of the algorithms and software that optimize the use of the new networks, and the design of the methodology and software that accurately measure the performance of the new networks. This project not only provides exciting hands-on education opportunities for students, but it also benefits the scientific community by improving the efficiency of scientific computational tasks. Optical circuit switching technologies have negligible power consumption with data-rate agnostic properties, thus have the potential to overcome the inherent challenges of conventional network architectures consisting of power-hungry electrical packet switches that must be upgraded frequently. Furthermore, optical circuit switching enables tuning of the network topology by dynamically reconfiguring the circuits at runtime, leading to high performance. However, approaches for achieving tunability in optical network architectures are nascent and the community lacks standard performance benchmarks. This project develops a new family of tunable optical network architectures using optical switches at the network edge to support diverse applications with high scalability, reliability, and low control-plane overhead, and develops standard benchmarks along with general methodologies to quantitatively compare different optical network architectures based on diverse performance metrics. Further, this project aims to cover a broad range of evaluations such as workload-independent analysis, large-scale simulation, and prototype implementation with real applications for ensuring the practical viability of novel optical network architectures. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.