Parallel Computing Using Optical Interconnections

Abstract

The advances in optical technologies have made it possible to implement optical interconnections in future MPP systems. This book is motivated by the fact that the area of massively parallel processing using optical interconnections has received increased attention in last few years, and noticeable progress has been made. The book is well organized to provide current and comprehensive coverage of the field, reflects the state-of-the-art from high-level architecture design and algorithmic points of view, and points out directions for further research and development.

The book is organized with two parts, with six chapters in each part. Part one covers optical interconnection networks and system architectures. Chapter 1, written by D. C. Hoffmeister, J. Chu, J. A. Perreault, and P. Dowd, provides a reconfigurable, hierarchical, wavelength division multiplexing (WDM) network called {Lightning for large-scale parallel computing in a distributed shared memory environment. Chapter 2, contributed by T. H. Szymanski, examines the class of Intelligent Optical Networks. In Chapter 3, A. Louri and B. Weech describe a scalable interconnection topology called Spanning Multichannel Linked Hypercube (SMLH) as well as its optical implementation, and compare SMLH with many popular networks. Chapter 4, by R. Melhem, G. Gravenstreter, D. Chiarulli, and S. Levitan, is devoted to the communication capabilities of the Partitioned Optical Passive Stars (POSP) network, together with message routing and scheduling on POSP, and analytical performance evaluation and simulations results. Chapter 5, authored by C.-F. Wang and S. Sahni, deals with the characteristics of parallel computers with optical transpose interconnection systems (OTIS), and summarizes the complexities of basic operations and fundamental algorithms on these systems. In Chapter 6, H.-A. Choi and E. J. Harder address the problems of routing and wavelength assignment in efficient use of WDM optical networks.

Part two of the book focuses on models and algorithms for optical interconnections. Several models are presented that make algorithm design, specification, and performance comparison relatively easier without worrying about optical engineering details. Algorithms and applications in this part include selection and sorting, numerical computations, matrix operations, image processing, and others. In Chapter 7, S. Q. Zheng proposes the hypernetwork model for optical interconnection networks, and shows how to design hypernetworks by using theory of hypergraphs and combinatorial block design. Chapter 8, written by C. Qiao, presents the Reconfigurable Array with Spanning Optical Buses (RASOB) model, together with example algorithms and embeddings Chapter contributed by S. Rajasekaran and S. Sahni, discusses and analyzes basic and important algorithms, deterministic and randomized, on the Array with Reconfigurable Optical Buses (AROB) model. In Chapter 10, S. D. Pavel and S. G. Akl develop optimal constant time algorithms for applications in image processing on the AROB model. Chapter 11, authored by Y. Pan, gives a comprehensive presentation of a model called Linear Array with Reconfigurable Pipelined Bus System (LARPBS), and implementation details of a number of primitive and fundamental operations. Finally, in Chapter 12, K. Li demonstrates novel implementations of fast and processor efficient parallel algorithms for matrix multiplication and related matrix manipulations on the LARPBS model by using the building blocks in Chapter 11.

This is the first book addressing parallel architectures, algorithms, and applications using optical interconnection networks. Then, this book can be an important reference book for anyone interested in of parallel/distributed computation using optical interconnections and especially for those beginners in the parallel/distributed computing field, which is based on optical interconnection networks. In addition, the book can be used as a text for a research oriented and/or seminar based advanced graduate courses, as well as a supplementary book for regular courses on Parallel Architectures, Parallel Algorithms, and High performance Computing.

Haklin Kimm
University of Tennessee in Martin