University of Liverpool runs Condor alongside PowerMAN
|
The University of Liverpool use the popular Condor platform for High Throughput Computing (HTC) tasks.
The Condor system, maintained by the University of Wisconsin, is a freely available toolset
which allows spare computing capacity to run user computations.
Dr. Ian Smith from the Liverpool Computing Services Department has published a paper explaining how Condor
has been successfully used alongside PowerMAN in a power-saving environment.
|
|
"Through careful monitoring and tailoring of the power management policy it has been
possible to remove around 200 000 - 250 000 hours of inactivity each week (resulting in an energy saving
of around 20-25 MWh based on an average consumption of 100 W per machine). This has led to an
estimated saving in electricity bills of around £124 000 per annum." - Dr. Ian Smith, University of Liverpool
Condor is used in a number of other universities and this paper, together with the associated presentation, provides
an excellent explanation of how PowerMAN can be effective in such an environment.
Abstract
The University of Liverpool Computing Services Department (CSD) Condor Pool has been developed over a
period of four years and now comprises approximately 300 classroom PCs (giving around 600 job slots) which
run the University’s Managed Windows (XP) Service. Many of our Condor users have eschewed writing their
applications in traditional programming languages such as C and FORTRAN in favour of using MATLAB M-files
compiled into "standalone" executables. Applications have varied widely from the simulation of avian influenza
transmission in poultry flocks to the modelling of the biological effects of radiation on normal tissue in
radiotherapy research. Whilst MATLAB provides users with a convenient environment for rapid development of
application codes, there are a number of hurdles to overcome in running MATLAB applications successfully under Condor.
MATLAB provides some unique benefits over traditional programming languages and both problems and merits are discussed here.
Since 2006, we have adopted a pro-active strategy to reducing IT hardware energy consumption through the adoption of
power-saving on PCs across campus. This has required us to adapt the Condor pool so that, firstly, Condor jobs are
not terminated early by PCs hibernating and, secondly, that power-saving PCs can be made available on-demand to run
Condor jobs. The approach taken to both issues is described here.
Full Paper
Experiences with Running MATLAB Applications on a Power-Saving Condor Pool
Related Links
Towards a greener Condor pool: adapting Condor for use with energy-efficient PCs
University of Liverpool Condor Service
University of Wisconsin Condor Homepage