Development and Feasibility Study of an Integrated High-speed Photonic Search Engine with Applications to Information Technology

University Of Arizona, Tucson AZ

Investigators

Abstract

Searching is a fundamental and pervasive operation in information processing. It can be expressed in a variety of forms such as (a) equivalence search (exact match search, not-equal-to search, similar-to search) (b) magnitude comparison search (smaller than, greater than, not-smaller-than, not-greater-than), (c) between limits search (search within a range or outside a range), (d) ordered retrieval search (ascending order, descending order), (e) extreme search (finding the maximum, the minimum, and the median), and (f) lookup search. These types of searches can be found in many information technology (IT) applications including database processing, artificial intelligence, pattern matching, security, and communications. In addition to its widespread use in information processing, searching plays a major role in data communication and routing where it serves as the basis for routing table lookup and classification necessary for packet routing. Conventional processors are often designed for numerical processing and search operations are delegated to software. Consequently, and for many applications, search operations are the most time consuming and create a major performance bottleneck. In this research project, we propose to investigate the development and detailed analysis of novel photonic search engines, that can implement search operations in hardware as directly as possible, and thereby as efficiently as possible. This will not only provide for minimum execution time but will also increase efficiency in other areas. These engines can be used as co-processing assists within a larger electronic processing node for computation and/or communication or as stand-alone search accelerators. We propose a novel approach that will combine device technology with architectural innovations in an efficient and practical manner with the aim of achieving a significant improvement in performance as well as reduction in system size. Our approach differs significantly from any on-going research on the subject. It represents a paradigm shift where search functions are directly executed in hardware instead of coded as lengthy subroutines using low-level instructions. It is anticipated that this approach will provide parallel search engines with low-power operation, high-speed processing (in the Terabits/second range), high degree of parallelism (due to the exploitation of several degrees of freedom in optics), and small monolithically integrated system size. The proposed research will have a broader impact on information technology. It is likely that a new field of study, multi-wavelength guided-wave integrated optical processing, will emerge for designing smaller processing systems while providing even more computing power than before.