An Efficient Priority-Aware Scheduling Approach for Federated Datacenter Scheduling Frameworks

Cloud computing serves as a cornerstone of modern IT infrastructure, delivering on-demand storage, computation, and networking resources. While various scheduling approaches have evolved to manage these resources, federated scheduling represents the latest advancement, addressing limitations of previous methods through a layered architecture with distributed scheduling nodes. This paper enhances the Megha scheduler by introducing a priority-based mechanism specifically optimized for latency-sensitive short-term jobs. We present four scheduling strategies-static allocation, repartitioning, dynamic allocation, and preemption-based scheduling-and evaluate their performance against the original Megha scheduler. Our simulation results demonstrate that the preemption-based approach delivers the most balanced performance improvement, reducing shortterm job latency by up to 99.9 percent while maintaining acceptable performance for long-term jobs. These findings highlight how priority-awareness in federated schedulers can dramatically improve responsiveness for time-sensitive applications without significantly compromising overall datacenter efficiency.