How can parallel job scheduling be optimized for dynamic environments? This research addresses the dynamic parallel job scheduling challenge by utilizing Gang scheduling. Aiming to enhance traditional Gang scheduling techniques, the study introduces Concurrent Gang, a class of scheduling policies that allow flexible and concurrent scheduling of multiple parallel jobs. This improves Gang scheduling's space sharing features while preserving its other benefits. To evaluate Concurrent Gang performance, the paper presents a dynamic competitive ratio methodology, which is then used to compare dynamic bin packing algorithms applied to a workload generated by a statistical model. Moreover, dynamic competitive ratio is the figure of merit used to evaluate and compare packing strategies for job scheduling under multiple constraints. The results show that in unidimensional cases, the difference between best fit and first fit is small, allowing first fit to be used without system degradation. The work suggests that packing algorithms must balance resource utilization across all dimensions when considering multidimensional cases, like memory, rather than prioritizing a single dimension. It provides guidelines for optimizing job scheduling under multiple constraints.
Published in the International Journal of Foundations of Computer Science, this article aligns with the journal’s focus on theoretical foundations of computer science. It relates to job scheduling algorithms, contributing to topics such as distributed computing and resource management, key interests of the journal.