[Starpu-devel] StarPU for distributed computing

[Anders Sjögren <anders.sjogren@chalmers.se>]

Dear StarPU developers.

I am currently in the situation where I am about to implement a framework for distributing a certain type of data analysis methods. I'd like the implementation to make use of distributed computing, as well as heterogeneous compute nodes. 

If I've understood correctly, StarPU solely handles the latter case, i.e. heterogeneous compute nodes, and not the distributed computing layer? 

Is there any ongoing work on taking StarPU to the next level, forming a runtime scheduler of distributed heterogenous resources?

Could you give some advice on how StarPU is typically used in distributed applications, when a high-level view such as algorithmic skeletons is preferred? 

It seems one would need to combine StarPU with a library with capabilities for distributed computing (e.g. Muesli from the group in Münster).

Best regards

Anders Sjögren

-------------------------------------------------------------

Anders Sjögren, PhD

Postdoctoral Researcher
Department of Mathematical Statistics
Chalmers University of Technology, Sweden

Email: anders.sjogren@chalmers.se

Phone: +46 (0)31 772 5315

-------------------------------------------------------------