Abstract
The work we present in this paper focuses on dynamic provisioning of computational resources depending on the performance requirements of the application and the characteristics of the cluster available for execution. We target applications which exhibit variable performance over time. The idea is to dynamically optimize the cost (total CPU-time) / performance (program execution time) ratio of parallel applications by (1) reducing the number of processors when the computation requirements decrease enough to justify it, and (2) moving computation onto those processors that can compute faster but don’t require extensive remapping of the data. This approach adapts well to time-shared platforms in which many applications may need to execute on the same cluster at the same time, and allows users to implement different cost / performance tradeoffs.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Huang, C., Lawlor, O., Kale, L.: Adaptive MPI. In: Rauchwerger, L. (ed.) LCPC 2003. LNCS, vol. 2958, pp. 306–322. Springer, Heidelberg (2004)
Martín, G., et al.: Runtime support for elastic execution of epigraph. Technical report (2012)
Martín, G., Marinescu, M., Singh, D., Carretero, J.: Leveraging social networks for understanding the evolution of epidemics. BMC Syst. Biol. 5(3) (2011)
Weatherly, D., Lowenthal, D., Nakazawa, M., Lowenthal, F.: Dyn-mpi: Supporting mpi on medium-scale, non-dedicated clusters. JPDC 2006 66(6), 822–838 (2006)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2012 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Martín, G., Singh, D.E., Marinescu, MC., Carretero, J. (2012). Runtime Support for Adaptive Resource Provisioning in MPI Applications. In: Träff, J.L., Benkner, S., Dongarra, J.J. (eds) Recent Advances in the Message Passing Interface. EuroMPI 2012. Lecture Notes in Computer Science, vol 7490. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-33518-1_38
Download citation
DOI: https://doi.org/10.1007/978-3-642-33518-1_38
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-33517-4
Online ISBN: 978-3-642-33518-1
eBook Packages: Computer ScienceComputer Science (R0)