An Evaluation of Optimizing for FUD in Scheduling for Shared Computing Environments

Scheduling systems for shared computing environments such as Cloud, Grid and Cluster computing act as the intermediator agent between the participating users with their requests in form of jobs or tasks, and the resources being shared. We will look at nine scheduling systems in this paper, and evaluate how each would be measured against the three optimizing parameters of Fairness, Utilization and Dynamicity (FUD). The variables are independent to each other, but design decisions made to optimize one variable affects the other two variables. We will demonstrate that batch job-based scheduling systems tend to focus on utilization as the means to reduce response time, whereas interactive task-based scheduling systems tend to focus more on dynamicity as the means to deal with rapid changes in the environment.