Would you run it here... or there? (AHS: Automatic heterogeneous supercomputing)

Parallel programs often exhibit strong preferences for different system structures, and machines with the ideal structures may all be available within a single heterogeneous network. There is also the complication that, although a particular application might execute fastest -when running by itself on one system, the best turnaround time might result from running the program on a different system that is less heavily loaded at the time the job is submitted. This paper takes a very practical approach to the problem of automatically making efficient use of heterogeneous supercomputing. Rather than making a heroic effort to achieve near peak speeds on a particular machine, our system (AHS) attempts to invisibly seek out and use whatever hardware will make the user's program execute fastest. Both the theory and implementation of AHS are presented.