Revenue maximisation for scheduling deadline-constrained mouldable jobs on high performance computing as a service platforms
Kuo Chan Huang, Chun-Hao Hung, Wei Hsieh · International Journal of High Performance Computing and Networking · 2017
Traditionally, high-performance computing (HPC) systems usually deal with the so-called best-effort jobs which do not have deadlines and are scheduled in an as-quick-as-possible manner. Recently the concept of HPC as a service (HPCaaS) was proposed, aiming to transform HPC facilities and applications into a more convenient and accessible service model. To achieve that goal, there will be new issues to explore, such as scheduling jobs with deadlines and maximising the revenue of service providers. This paper presents a reservation-based dynamic scheduling approach for scheduling deadline-constrained mouldable jobs with the aim of maximising a service provider's revenue. The proposed approach has been evaluated with a series of simulation experiments. The experimental results indicate that our scheduling approach can achieve significantly higher revenue than previous methods. In the experiments, we also explored several research issues, including waiting queue sequencing, processor allocation decisions on time and space, admission control, and partial rescheduling.