Overview
One approach to building supercomputers is to interconnect any number of common PCs using an interface like Ethernet. This method, called clustering or "Beowulf" clustering, produces very inexpensive powerful computers whose capabilities would previously have cost millions of dollars. Now for a few tens of thousands of dollars, very powerful computers can be built for any number of computing solutions. From scientific applications to transaction processing, clustering technology provides an affordable, scalable computing solution.
Building Linux Clusters introduces the reader to the basics of cluster installation and configuration, and comes complete with a CD full of cluster installation programs and tools for parallel programming. Focusing on the "how to" of building a Linux cluster, this book is a hands-on guide for people new to clustering. It is the definitive guide to scaling Linux for scientific and enterprise applications.
A guide for experienced Linux users showcases the operating system's ability to connect or "cluster" microcomputers that will then mimic the performance of supercomputers, and includes tips on installation, customizing, and parallel programming
Editorials
From Barnes & Noble
The Barnes & Noble ReviewSupercomputing is no longer limited to folks modeling nuclear weapons explosions. Thanks to Linux and clustering, you can build a working supercomputer for a tiny fraction of what it used to cost Uncle Sam, then use it for whatever real-world enterprise or scientific applications you can dream up. It's not quite a spare time/spare change project yet, but it's getting there. Curious? Serious? Either way, there's now a hands-on, comprehensive guide to Linux clustering: Building Linux Clusters.
David Spector introduces cluster installation and configuration in depth, explaining the role of networking and TCP/IP; key design considerations; how to choose CPUs, disks, and networking approaches; providing for data access, messaging and queueing; and more. You'll walk through cluster-building, start-to-finish: preparing your environment, building nodes, installation, cabling, software configuration, day-to-day cluster management, and more. You'll also find detailed coverage of tools, libraries, and techniques for Linux-based parallel programming.
Finally, Spector profiles three applications that take supercomputing out of the lab: PVMPOV, a parallel version of Persistence of Vision (POV), a popular ray-tracing package for computer-generated animation; PVFS, which allows you to construct a supercharged filesystem out of a Linux cluster; even mp3pvm, for RIPping MP3 files in one heck of a hurry.
—Bill Camarda, bn.com editor