next up previous contents index
Next: Overview Up: Cplant Parallel Application Support: Previous: Contents   Contents   Index

Introduction

This document describes the utilities, daemons, configuration files, and scripts in the Cplant runtime environment that combine to do the following

It also describes how PBS (Portable Batch System) works on Cplant, including how to install it, configure it, run it, and handle common problems that may occur.

It is intended for administrators who will be setting up a Cplant and maintaining it while it is running. This document is targeted to the Sandia National Laboratories site in Albuquerque. There may be differences at other sites, such as location of configuration files. Other Cplant sites are encouraged to write a site guide to supplement this guide.

It is assumed that the reader is familiar with the Cplant architecture and in particular

We frequently refer to nodes in this document and node numbers. The workstations in the cluster which are playing the role of service node or compute node are the nodes. The service nodes host users loading parallel applications and querying for job status, among other things. The compute nodes each host a single process of a parallel application. The node numbers we refer to are indices into a map of Myrinet hardware addresses, are contiguous, and begin at zero. These are the numbers displayed with a label like Node ID in the status displays and error messages of yod and pingd .

Other helpful information may be found at the Cplant web site,
[4]http://www.cs.sandia.gov/cplant or by email queries to
[4]alaska-help@cs.sandia.gov.

This document describes revision number 1.0 of the Cplant code.


next up previous contents index
Next: Overview Up: Cplant Parallel Application Support: Previous: Contents   Contents   Index
Lee Ann Fisk 2001-06-25