Subjects: Computer Science >> Other Disciplines of Computer Science Subjects: Physics >> Condensed Matter: Electronic Structure, Electrical, Magnetic, and Optical Properties submitted time 2022-08-15
Abstract: With the development of high-performance computing architectures, many software and hardware have a multi-layer parallel structure. A large amount of allocation schemes can be involved when users allocate multi-layered system resources to many computational tasks distributed in different vertical tiers and horizontal groupings. It is becoming increasingly difficult for users to determine the optimal parallel parameters and hardware resource usage. We investigate an optimization method which is helpful for users to automate the determination of the optimal application parallel parameters and hardware usage for high-efficient and/or large-scale computation. In addition, we propose a solution that deeply integrates the optimization method with the job scheduling system, which has produced excellent practical results.
Peer Review Status:Awaiting Review