Cluster computing: the state-of-the-art in theory and practice Rapid improvements in network and processor performance are revolutionizing high-performance computing, transforming clustered commodity workstations into the supercomputing solution of choice. This book brings together contributions from more than 100 leading practitioners, offering a single source for up-to-the-minute information on virtually every key system-related issue in high-performance cluster computing. The book contains expert coverage of commodity supercomputing systems and architectures; Internet-based wide area metacomputing systems; the role of Java; new applications and algorithms; advanced techniques for enhancing availability and throughput; and much more. Discover the state-of-the-art in: * Communal multiprocessing/adaptive parallelism techniques for resource sharing * Networking, lightweight protocols, active messages, killer switches, and I/O * Cluster middleware and resource management systems * Cluster computing programming environments, tools, and paradigms * Administering high-performance clustered systems High Performance Cluster Computing, Volume 1: Architectures and Systems captures the
High Performance Cluster Computing contains academic articles concerning supercomputing collected from researchers around the world. Though targeted primarily at graduate students and researchers in computer science, the general reader may find great value in its overview of the current state of high-performance computing.
Computer science experts address many aspects of high performance computing, beginning with the state-of-the-art concepts and basic terminology related to cluster computing. Their investigations provide immediate solutions to engineering problems like optimized node arrangements for low-cost workstations yoked together to solve problems in parallel. One article describes such a cluster created for the Department of Energy that uses 9,000 Pentium CPUs to model nuclear detonations.
Various contributors also consider the requirements necessary for improving parallel programs in terms of speed and logic, including reductions in network latencies and enhanced file and I/O access. One contributor even suggests that Network RAM--unused RAM in systems on the same network--may someday challenge the hard disk for fast--and permanent--data storage.
In all, High Performance Cluster Computing works as an up-to-date, central repository of current thinking on interconnecting computers and processors to improve speed and performance. It provides a valuable roadmap of the state of the art in computer science research as well as some potential benefits for forward-looking corporate computing professionals. --Richard Dragan
