Comparing the effectiveness of workload balancing objectives in FMS loading

Workload balancing is a popular objective in FMS loading. Usually it is achieved by minimizing the imbalance, which is some function of workload differences. Thus, each imbalance measure provides a distinct balancing objective. The literature provides many balancing objectives but very few comparative studies. In these studies, the various balancing objectives are compared based on their correlation with the ultimate performance criteria like throughput and makespan. The present study has a different motivation. It compares the balancing objectives based on their effectiveness for balancing itself. Nine balancing objectives are considered. Some loading problems are generated at three utilization levels. For our inferences, a summary value was devised, efficacy, that is the number of other balancing objectives dominated by a particular balancing objective. An interesting observation is that high utilization itself acts as a balancing agent, making a balancing objective less effective than at medium or low utilization. Individually, 'min [avg. pairwise difference]' is the most effective objective, followed by 'min [max pairwise difference]'. Thus, the 'pairwise consideration' of workloads emerges as the best theme for imbalance measurement. The results are in agreement with the previous findings.