1.1-6 Morrison Hotel
Monitoring and Testing |
||
Back to Main Documentation Page |
There are three ways to centrally monitor the cluster. Each has a different information "focus" which is useful in different ways.
Note: WWTop does not give queue information, but is updated more quickly than Userstat. It also gives Ethernet network statistics.
Having baseline performance numbers for cluster performance is important for two reasons. First, it provides confidence that the cluster is running correctly. And, second, it provides a basis for which to measure changes due to hardware and software upgrades. It is preferable that upgrades provide performance as good as or better than before, but this is not always the case.
In order to address these issues and to make it easy to run tests and see the results, the Beowulf Performance Suite (BPS) was created. The BPS should be run under a user account (not root). When finished you can create an HTML page with results like this page. There is a man page (man bps) that will help explain its use. An on-line article called A Tool for Cluster Performance Tuning and Optimization maybe helpful as well. When finished running the tests, you should be able to link the HTML results for your cluster to the Main Documentation Page.
This page, and all contents, are Copyright (c) 2007-2013 by Basement Supercomputing, Bethlehem, PA, USA, All Rights Reserved. This notice must appear on all copies (electronic, paper, or otherwise) of this document.