Slide 15 of 47
Notes:
50 Scenarios Executed (50 Results Pictured)
The average slowdown over all executions for the round-robin strategy was 18.835. The average slowdown over all executions for the Enhanced PVM Strategy was 11.698. On the real cluster, the Enhanced PVM Strategy cut 38% off the average slowdown.
We picked our simulation parameters, e.g. ?, conservatively, to minimize the advantage of the Enhanced PVM Strategy and our other strategies. This is why our strategy performed better in the real world than in the simulations.
