AI Based Workflow Scheduling and Automatic Migration Planning Model for Cloud Computing

Scheduling tasks in information technology facilities in an energy-efficient manner is an important and widely-discussed problem. But in a real-world information center, the job execution times are all over the place and difficult to predict. The fact that scheduling choices based on task arrival timings are likely to lead to wasted energy or resources that remain idle over time had been noticeably ignored by previous studies. This work proposes a new energy-efficient task scheduling method that addresses these shortcomings by combining the concepts of assignment as well as migrating for jobs with varying durations. Transferring long-running processes to low-load servers necessitates assigned tasks, which in turn improves resource efficiency. Specifically: (1) Organizing assignments with varying durations is recast as an MDP issue with large-scale parameters; (2) Our solution to this massive MDP challenge is a well-designed Deep Reinforcement Learning (DRL) system for assignment as well as migration decision-making. Various enhancements are suggested to accomplish online training in order to render the DRL algorithms more applicable in real-world situations; (3) Using actual data, we were able to demonstrate that our approach achieves 14% energy savings compared to the current baselines, all while maintaining the identical degree of QoS.