Scheduling Critical Periodic Jobs with Partial Computations along with Gang Jobs
Helen D. Karatza · 2023
One of the key issues in distributed systems, such as clouds, is scheduling of complex workloads which consist of jobs with different characteristics. One type of parallel jobs running in these systems are gang jobs. In this paper we study scheduling of workloads consisting of gangs as well as of critical periodic jobs which can accommodate partial computations when they need to overcome the execution of gang jobs. Two different types of workloads are examined based on the gang applications characteristics. The simulation results provide useful insights into how the partial computations of critical jobs affect the performance of gangs. The results also demonstrate that partial computations, under the proposed scheduling scheme, can reduce the mean response time of the gangs jobs, incurring only an insignificant degradation in their average result precision.