Scyld ClusterWare HPC

User's Guide

Table of Contents
Scyld ClusterWare Overview
What Is a Beowulf Cluster?
A Brief History of the Beowulf
First-Generation Beowulf Clusters
Scyld ClusterWare: A New Generation of Beowulf
Scyld ClusterWare Technical Summary
Top-Level Features of Scyld ClusterWare
Process Space Migration Technology
Compute Node Provisioning
Compute Node Categories
Compute Node States
Major Software Components
Typical Applications of Scyld ClusterWare
Interacting With the System
Verifying the Availability of Nodes
Monitoring Node Status
The BeoStatus GUI Tool
The bpstat Command Line Tool
The beostat Command Line Tool
Issuing Commands
Commands on the Master Node
Commands on the Compute Node
Copying Data to the Compute Nodes
Sharing Data via NFS
Copying Data via bpcp
Programmatic Data Transfer
Data Transfer by Migration
Monitoring and Controlling Processes
Running Programs
Program Execution Concepts
Stand-Alone Computer vs. Scyld Cluster
Traditional Beowulf Cluster vs. Scyld Cluster
Program Execution Examples
Running Programs That Aren't Parallelized
Starting and Migrating Programs to Compute Nodes (bpsh)
Copying Information to Compute Nodes (bpcp)
Running Parallel Programs
An Introduction to Parallel Programming APIs
Mapping Jobs to Compute Nodes
Running MPICH Programs
Running OpenMPI Programs
Running PVM-Aware Programs
Porting Other Parallelized Programs
Running Serial Programs in Parallel
Job Batching
Job Batching Options for ClusterWare
Job Batching with TORQUE
File Systems
Sample Programs Included with Scyld ClusterWare
Glossary of Parallel Computing Terms