How to Build A Beowulf

The early supercomputers used distributed computing and parallel processing to link processors together in a single machine, often called a mainframe. Exploiting the same technology, researchers are now using off-the-shelf PCs to produce computers with supercomputer performance. It is now possible to make a supercomputer for less than $40,000. Given this new affordability, a number of universities and research laboratories are experimenting with installing such Beowulf-type systems in their facilities.

This how-to guide provides step-by-step instructions for building a Beowulf-type computer, including the physical elements that make up a clustered PC computing system, the software required (most of which is freely available), and insights on how to organize the code to exploit parallelism. The book also includes a list of potential pitfalls.

Table of contents

Series Foreword

Foreword

Preface

1: Introduction

1.1: A Brief History
1.2: The Beowulf Book

2: Overview of Beowulf Systems

2.1: What Is a Beowulf?
2.2: A Taxonomy of Parallel Computing
2.3: Benefits of Beowulf
2.4: A Critical Technology Convergence
2.5: The Beowulf System Node
2.6: The Beowulf Network
2.7: Linux
2.8: Message Passing for Interprocessor Communication
2.9: Beowulf System Management
2.10: The Beowulf Challenge

3: Node Hardware

3.1: Overview of a Beowulf Node
3.2: Processors
3.3: Motherboard
3.4: Memory
3.5: BIOS
3.6: Secondary Storage
3.7: PCI Bus
3.8: Examples of a Beowulf Node
3.9: Boxes, Shelves, Piles, and Racks
3.10: Node Assembly

4: The Linux Operating System

4.1: History of Linux
4.2: Linux Kernels and Linux Distributions
4.3: Linux Features
4.4: File Systems
4.5: System Configuration
4.6: Tools for Program Development
4.7: Linux's Unique Features
4.8: Installing an Initial System
4.9: Keeping Up with Linux
4.10: Suggested References

5: Network Hardware and Software

5.1: Fast Ethernet
5.2: Alternative Network Technologies
5.3: TCP/IP
5.4: Sockets
5.5: Higher Level Protocols
5.6: Distributed File Systems
5.7: Remote Command Execution

6: Managing Ensembles

6.1: System Access Models
6.2: Assigning Names
6.3: Cloning Nodes
6.4: Basic System Administration
6.5: Defending the Pack: Security Strategies
6.6: Job Scheduling

7: Parallel Applications

7.1: Parallelism
7.2: Broad Categories of Parallel Algorithms
7.3: Process-level Parallelism

8: MPI - A User-level Message-passing Interface

8.1: History
8.2: MPI Basic Functionality
8.3: Parallel Data Structures with MPI
8.4: MPI Advanced Features

9: Programming with MPI - A Detailed Example

9.1: Example: Sorting a List of Uniformly Distributed Integers
9.2: Analysis of Integer Sort
9.3: Measurement of Integer Sort
9.4: Example: Sorting with User-supplied Comparator
9.5: Analysis of a More General Sort
9.6: Summary

10: Conclusions and Views

10.1: New Generation Beowulfs
10.2: New Opportunities
10.3: True Costs
10.4: Total Work versus Peak Performance
10.5: Big Memory versus Out-of-core
10.6: Bounding Influences on the Beowulf Domain
10.7: New Programming Models
10.8: Will Linux Survive the Mass Market?
10.9: Final Thoughts

Index of Acronyms

Index