Study on Locality, Fairness, and Optimal Resource Allocation in Cluster Scheduling

In the field of computing, evaluating system performance is critical. The need for new scheduling paradigms, particularly in the context of large cluster computing systems (Lu et al. in IEEE Trans Parallel Distrib Syst 34:1145–1158, 2023), is paramount. The authors will examine the feasibility of multiple techniques that strive to impart a fair distribution of resources across systems, with the endgame being to minimize latency. Moreover, the authors will explore the feasibility of devising solutions with varying degrees of enhanced data locality, which has consistently been the backbone of efficient scheduling. Another direction of the study will investigate the advantages and disadvantages of two contrasting approaches: the first of these will look into the efficacy of erasing specific jobs and freeing up resources for the next set of tasks, whereas the alternative will attempt to finish pending jobs before tackling the remaining assignments. Finally, the authors conclude that they abided by professional norms in extracting and analyzing the results of their experiments (Wennmann et al. in Investig Radiol 58(4):253–264, 2023).