Optimisation of job scheduling for supercomputers with burst buffers

The ever-increasing gap between compute and I/O performance in HPC platforms, together with the development of novel NVMe storage devices (NVRAM), led to the emergence of the burst buffer concept - an intermediate persistent storage layer logically positioned between random-access main memory and a parallel file system. Since the appearance of this technology, numerous supercomputers have been equipped with burst buffers exploring various architectures. Despite the development of real-world architectures as well as research concepts, Resource and Job Management Systems, such as Slurm, provide only marginal support for scheduling jobs with burst buffer requirements. This research is primarily motivated by the alerting observation that burst buffers are omitted from reservations in the procedure of backfilling in existing job schedulers. In this dissertation, we forge a detailed supercomputer simulator based on Batsim and SimGrid, which is capable of simulating I/O contention and I/O congestion effects. Due to the lack of publicly available workloads with burst buffer requests, we create a burst buffer request distribution model derived from Parallel Workload Archive logs. We investigate the impact of burst buffer reservations on the overall efficiency of online job scheduling for canonical algorithms: First-Come-First-Served (FCFS) and Shortest-Job-First (SJF) EASY-backfilling. Our results indicate that the lack of burst buffer reservations in backfilling may significantly deteriorate the performance of scheduling. [...] Furthermore, this lack of reservations may cause the starvation of medium-size and wide jobs. Finally, we propose a burst-buffer-aware plan-based scheduling algorithm with simulated annealing optimisation, which improves the mean waiting time by over 20% and mean bounded slowdown by 27% compared to the SJF EASY-backfilling.