Preface

Welcome to the Scyld ClusterWare HPC Administrator's Guide. It is written for use by Scyld ClusterWare administrators and advanced users. This document covers cluster configuration, maintenance, and optimization. As is typical for any Linux-based system, the administrator must have root privileges to perform the administrative tasks described in this document.

The beginning of this guide describes the Scyld ClusterWare system architecture and design; it is critical to understand this information in order to properly configure and administer the cluster. The guide then provides specific information about tools and methods for setting up and maintaining the cluster, the cluster boot process, ways to control cluster usage, methods for batching jobs and controlling the job queue, how load balancing is handled in the cluster, and optional tools that can be useful in administrating your cluster. Finally, the an appendix covers the important files and directories that pertain to operation of Scyld ClusterWare.

This guide is written with the assumption that the administrator has a background in a Unix or Linux operating environment; therefore, the document does not cover basic Linux system administration. If you do not have sufficient knowledge for using or administering a Linux system, we recommend that you first consult Linux in a Nutshell and other useful books published by O'Reilly and Associates.

When appropriate, this document refers the reader to other parts of the Scyld documentation set for more detailed explanations of the topic at hand. For information on the initial installation of Scyld ClusterWare, refer to the Installation Guide, which provides explicit detail on setting up the master node and booting the compute nodes. For administrators who are new to the ClusterWare concept, we recommend reading the User's Guide first, as it introduces ClusterWare computing concepts.