C-141 HomeDocumentsSoftwarePDS HomeHPCLab

Outline

During the last decade the proposition of grids has influenced many aspects of contemporary high performance computing. Nowadays grids are entering a new phase where one can observe increasing deployment of grids, advances on the afforded middleware, and broadening of the group of target applications that are considered appropriate for execution on top of grids. Nevertheless there are many open issues that have to be addressed in order to move towards the vision of grids as providers of computational power. Many of these issues are strongly related to the programming models that are available for grid programming and the trade-off that they succeed between simplicity and efficiency.

In the current project we will investigate the potentiality of a multithreaded shared memory programming model on top of grids. At the moment mainly embarrassingly parallel, or fully distributed applications with few dependencies are executing successfully on grids on a batch job fashion. Software tools will be implemented and integrated into the state of the art grid middleware and conclusions will be drawn concerning their effectiveness and their applicability.


Details

Starting date of the project: 01 May 2007


People Involved

Elefterios D. Polychronopoulos (Faculty)
Konstantinos I. Karantasis (PhD student)