bobMonitor (c) Robin Humble 2003-2012
License: GPLv3+
http://code.google.com/p/bob-monitor/


INTRODUCTION:

bobMonitor is a cluster montoring tool that presents near real-time job and cluster
information in a dynamic web page. it is aimed at both
cluster administrators and users, and enables them to visually see how jobs
and the cluster is performing.

python is used on the server side to aggregate and collate information and
generate an XML file of the cluster and job state.
typically data is harvested from PBS and ganglia and instantaneous and historical data for each job is computed.

JavaScript is used on the client-side (in the browser) to frequently update the cluster display.
historical data over the length of the job's run is visible.

bobMonitor has been developed since 2003 and should be portable to most ganglia enabled and PBS compatible cluster environments. it has been developed with HPC clusters and jobs in mind.


CONFIGURATION:

a single simple config file is used to setup preferences for both the server-side python code and the client-side JavaScript code.


REQUIREMENTS:

bobMonitor relies upon the following:
 - ganglia  http://ganglia.info/
 - a PBS variant such as OpenPBS or torque
 - python 2.4 or above (but not 3.*)
 - a web server with cgi-bin access
 - a JavaScript enabled web browser

server-side pie charts of the queue require:
 - PyChart  http://home.gna.org/pychart/ 
 - netpbm tools (netpbm-progs) for pieChart image post-processing

optional:
 - maui or moab scheduler

bobMonitor gratefully uses and includes a copy of Erik Arvidsson's
slider implementation:
  http://webfx.eae.net/dhtml/slider/slider.html
and a progress bar from:
  http://www.dynamicdrive.com/dynamicindex11/dhtmlprogress2.htm
  by Brian Gosselin of http://scriptasylum.com


INSTALLATION:

see the INSTALL file


TESTING:

RedHat 4/5 variants are the only tested server-side configuration.

most client testing has been done with gecko based browsers (eg. firefox), but
webkit (eg. safari, Android browser) has also been somewhat tested. IE is untested, but might work.


OPTIMISATIONS:

the bobMonitor client-side JavaScript code is optimised to lower the load on
browsers by updating only those fields that have changed since the last update.
it has been tested with clusters of 1500 nodes and it is expected to scale
to 10000+ node clusters (if your monitor is big enough!).
the ever improving JavaScript implementations in modern browsers will also help.

bandwidth required by the compressed XML stream is relatively small, but if you
have a large cluster probably shouldn't be attempted frequently over dial-up or
mobile.

the server-side load has been optimised by decoupling the state gathering and
aggregation from the web requests. typically the XML state of
the cluster is updated every 5 to 10 seconds or so, wheras the cgi-bin script
may send this XML stream to many clients every second.