This paper empirically investigates the effect of Grid Computing on the computational cost of Evolution Strategies (ES). In particular, in order to solve the dynamic task scheduling problem that arises in distributing the evolutionary computation across a Grid, we adopt List Scheduling with Round-robin order Replication (RR). Computer experiments using a desktop grid environment in a local network are conducted which compare the theoretical maximum energy loss with the actual measurement for a given scheduling task. We show that Grid Computing cuts the computational time of ES. In addition, the results of applying RR to the local network show that the actual energy loss is less than the theoretical maximum value.
- List Scheduling with Round-robin order Replication
- Evolution Strategies
- Grid Computing