Performance Modeling of Optoelectronic Communication Networks for Ultracompact Massively Parallel Processing Systems

点击翻译按钮获取中文摘要

9633673 Cruz-Rivera The objective of this Minority Research Planning Grant proposal is to present a plan for preliminary studies that will facilitate the development of a competitive research project in the area of optoelectronic communication networks, with the full expectation that this project will form the basis of a subsequent regular research proposal to the NSF's Optical Communication System program. In addition to the technical activities that will be carried out during the planning period, the work proposed herein seeks to address certain infrastructure-related issues that are unique to the establishment of an Optoelectronic Systems Research Group at the University of Puerto Rico-Mayaguez - a primarily undergraduate institution that is currently making great strides towards the development of a strong Ph.D. program in Computer Science and Engineering. The technical activities to be carried out during the proposed twelve month planning represent a single component of a broader area of investigation that would constitute the main focus of the Optoelectronic Systems Research Group at the University of Puerto Rico-Mayaguez (UPRM). In particular, the proposed line of inquiry seeks to advance the development of an Applications-Driven Optical Interconnect Technology (ADROIT) Model that would allow, for a given application class, the determination of : (1) the optimal optoelectronic system implementation given set of algorithmic, architectural, topological, and technological parameters, and (2) the relative impact the optical interconnect technologies in question would have on the overall cost/performance metric of these systems. The proposed activities would, on the one hand, advance the short-term development of certain aspects of this ADROIT model by providing insight into the topological-technological design space associated with the implementation of optoelectronic communication networks for ultracompact massively parallel processing sy stems, while on the other, provide a framework that will allow the identification of the specific research tasks to be undertaken and methodologies to be employed in the long-term development of this model. The identified tasks and methodologies would in turn form the basis for the subsequent research proposal. ***