HPC Scheduling Method Based on Debug and Background Job Classes Division

HPC users usually share supercomputer resources by submitting their jobs to the queue. The job management system scheduler forms job schedule for the supercomputer nodes. There are three widespread scheduling principles: assigning jobs fixed parameters such as number of nodes or cores and execution time, the priority non-preemptive ‘‘fair share’’ scheduling, and using one of backfilling strategies. The article presents the scheduling method based on these principles. The special feature of the proposed method is the division of the input job workflow into the debug, regular, and background jobs. The method is implemented as a part of the SUPPZ system for parallel job management, which has been used for more than 20 years in HPC systems at the Joint supercomputer center of the RAS (JSCC RAS) and the Keldysh Institute of Applied Mathematics of the RAS. The article contains the results of the JSCC RAS supercomputers statistics analysis of the proposed scheduling method effectiveness.