Design and Evaluation of Socially Just Scheduling Policies

Queueing for service occurs in technical applications as well as in common daily life, and scheduling should manage queues in a satisfactory way. Performance metrics are influenced by the service order, thus, scheduling policies should be designed with respect to the main performance metrics of interest. For human customers psychological effects and performance metrics other than delay play a crucial role in judging a service facility. Metrics related to social justice are of major importance, where individual perceptions of justice are strongly coupled with an equal treatment of customers. We quantify and measure social justice based on two types of customer discriminations: being overtaken and long waiting for single customers with large service requirements. In order to improve perceived justice of customers two scheduling policies, alternating FCFS/SJF and MFD (Most Frequently Discriminated), are proposed and shown to be useful. The prize that has to be paid to improve justice oriented queue metrics at the cost of higher expected delay or higher delay variability is examined by comparing the potential increases in expected delay and delay variability to the potential increase of social justice. A discrete event simulation study is provided for queueing systems operating under the traditional FCFS and SJF policies and under the new policies. Our results show that the new policies significantly improve social justice without too much worsening the delay metrics and that MFD even reduces the delay variability. Key–Words: Queueing, Scheduling, Social Justice, Discrimination Frequency, Discrete Event Simulation Study