Improving Performance via Computational Replication on a Large-Scale Computational Grid