Academic User Services' Beowulf

Overview

Dolly (As in the cloned sheep) will be the Beowulf cluster for the University of Oregon. Commodity PC hardware makes it possible for anyone to have their own supercomputer. Currently, planned uses are for rendering, software developement and familiarization with parallel environments and tools. We expect the initial installion to use the following hardware. All nine nodes will run Redhat 5.0, ORNL's pvm 3.3.11, and a number of beowulf specific extensions to the standard Linux Kernel.

This web page will provide programming resources necessary to the deployment of our Beowulf cluster.

Local Docuements about our Beowulf Project

• Current Hardware Specifications for our system.

• Notes on Building our System, a list of things we noticed were needed when we built our system, hopefully this will make it easier in the future. Currently this document is in pretty rough form.

• Disk Performance benchmarks. When we created the nfs server for our system we wanted to know if the software raid setup was really that fast (it was). These are the results and comparisons to some other landmark systems we have available.

More information about Beowulf

• NASA: Beowulf Links
  
• Beowulf Software
• PVM Home Page (ORNL)
  
• PVM Programming Info from Pittsburgh Supercomputing
  
• PVM Documentation from Cornell
  
• Draft of the Beowulf howto
• Building a Beowulf - caltech
• SGI: Models of Parallel Computation
• SGI: MPI and PVM User's Guide

• The Hive - NASA
• Hygliac & Naegling - Caltech
• Loki - LANL
• Topcat - University of Southern Queensland
• Megalon - Sandia