e-Science logo Nesc logo
 
 
About NeSC
e-Science Institute
e-Science Hub
TOE
Contacts
e-Science Events
Resources
Newsroom
Presentations & Lectures
Technical Papers
Global Grid Links
Projects
UK e-Science Centres
UK e-Science Teams
Career Opportunities
Bibliographic Database
 

 

Paper ID: 1732

Performance Prediction for a Code with Data-dependant Runtimes
Brian,Foley Paul,Isitt Daniel,Spooner Steven,Jarvis Graham,Nudd

Appeared in: Proceedings of the UK e-Science All Hands Conference 2005 website: http://www.allhands.org.uk/2005/
Page Numbers:
Publisher: Engineering and Physical Sciences Research Council
Year: 2005
ISBN/ISSN: 1-904425-53-4
Contributing Organisation(s):
Field of Science: e-Science

URL: http://www.allhands.org.uk/2005/proceedings/papers/527.pdf

Abstract: We describe a parallel medical imaging code which operates by successive refinement until a convergence test is passed. As such, its runtime is highly variable, 'data dependant', and isn't amenable to prediction by tools such as Warwick's PACE. We investigate the factors influencing the runtime of this code, and describe a means of estimating its runtime given specific input data. We also examine the use of 'speculative prediction', ie inserting tasks into the TITAN schedule that are scheduled but never run to provide the user with interactive feedback on the resource requirements of their workflows.

Keywords: e-Science, AHM 2005


BIB DOC HTM HTML PDF PPT PS RTF TEX TXT ZIP




 

Last Updated: 22 Jun 12 11:02
This is an archived website, preserved and hosted by the School of Physics and Astronomy at the University of Edinburgh. The School of Physics and Astronomy takes no responsibility for the content, accuracy or freshness of this website. Please email webmaster [at] ph [dot] ed [dot] ac [dot] uk for enquiries about this archive.