Groups.grid.org

Currently, the performance penalty on MPI jobs is pretty significant in the 'cloud'. (the communication layers...make it slow)

What wouldbe interesting to understand is if a solution that eased the degradation was available, would it be of value?

Comments welcome.