Towards Conflict-Aware Workload Co-execution on SX-Aurora TSUBASA

• Published in: Parallel and Distributed Computing, Applications and Technologies Vol. 13148; pp. 163 - 174
• Main Authors: Nunokawa, Riku, Shimomura, Yoichi, Agung, Mulya, Egawa, Ryusuke, Takizawa, Hiroyuki
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2022; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Performance interference; SX-Aurora TSUBASA; Workload colocation
• ISBN: 9783030967710; 3030967719
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-96772-7_16

NEC SX-Aurora TSUBASA is the latest vector supercomputer, consisting of host processors called Vector Hosts (VHs) and vector processors called Vector Engines (VEs). The final goal of this work is to simultaneously use both VHs and VEs to increase the resource utilization and improve the system throughput by co-executing more workloads. However, performance interferences among VH and VE workloads could occur because they share some computing resources and potentially compete to use the same resource at the same time, so-called resource conflicts. As the first step to achieve efficient workload co-execution, this paper experimentally investigates the performance interference between a VH and a VE, when each of the two processors executes a different workload. Our evaluation results clearly demonstrate that some characteristics of a workload such as system call frequency can be used as a good indicator to predict if the workload can affect the performance of another co-executing workload. We believe that this will be helpful to identify a pair of workloads causing frequent resource conflicts, and thus reduce the risk of performance interference between co-executing workloads on an SX-AT system.