Introduction

• The Dogwood Linux Cluster is a distributed memory multi-processor sy stem. Each node of the cluster contains two 3-GHz Intel Xeon EMT processors with 2048KB of cache (per cpu) and 3 GB of RAM (per node). The nodes are interconected with Gigabit Ethernet (60-11 0 Mbytes/sec bandwidth, 50-200 usecs latency).
• In the default queue, a maximum of 72 cpus may be used by a single u ser at at time, which can be spread over a maximum of 36 jobs. The maximum time permitted for a single job is 168 hours. A test queue (testq) is also available for testing and debugging purpo ses; this queue has a wall-time limit of one hour and a cpu limit of 4, which can be employed in up to 3 simultaneous jobs.
• The Dogwood Linux cluster uses ROCKS as its operating system and the Portable Batch System (P BS) software to distribute the computational workload across the nodes. PBS is a batch job scheduli ng application that provides the facility for building, submitting and processing both serial and par allel batch jobs on the cluster. The PBS User's Guide is posted online.
  

Useful Links

• Dogwood Cluster Status
  Ganglia Toolkit for displaying cluster load and status information.
• Dogwood Queue Status
  Displays jobs in the queue. Click on the link for a given job ID to get complete information about that job.
• PBS Script Generator
  Generates PBS scripts based on user input to a form.

PBS Documentation

• PBS Pro User's Guide