Don't hurry be green: scheduling servers shutdown in grid computing with deep reinforcement learning

Predicting the efficiency of job scheduler actions

“…Applying the correct strategy for job scheduling in a processing-intensive environment such as a HPC datacenter is no trivial task and the concerns in this choice range from environmental to financial. For example, choosing the most appropriate algorithm based on the profile and patterns of the workflow being processed will bring considerable energy usage benefits [Casagrande et al 2022]. In this context and with such objective, this work focused on using mathematical regression tools to obtain regression models consistently capable of predicting the outcome of SJF-, FCFS-, SAF-, EASY-, F1-, F2-, F3-and F4-scheduled workloads.…”