Jason Yalim
ABSTRACT
The popular SchedMD scheduler, Slurm, provides an internal algorithm for assigning priorities to user’s high-performance computing (HPC) jobs based on previous usage. This algorithm, aptly named fairshare, is classically an exponential function of a user’s usage history relative to the HPC population. This study explores an option HPC centers can take to increase the transparency of the classic fairshare algorithm and shows how usage and classic fairshare may be dynamically modeled using a simple differential equations approach.
- Y. Georgiou, D. Glesser, K. Rzadca, and D. Trystram. 2015. A Scheduler-Level Incentive Mechanism for Energy Efficiency in HPC. In 2015 15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing. 617–626.Google Scholar
- Andy B. Yoo, Morris A. Jette, and Mark Grondona. 2003. SLURM: Simple Linux Utility for Resource Management. In Job Scheduling Strategies for Parallel Processing, Dror Feitelson, Larry Rudolph, and Uwe Schwiegelshohn(Eds.). Springer Berlin Heidelberg, Berlin, Heidelberg, 44–60.Google Scholar
Index Terms
- Toward Dynamically Controlling Slurm’s Classic Fairshare Algorithm
Recommendations
Communication-aware Job Scheduling using SLURM
ICPP Workshops '20: Workshop Proceedings of the 49th International Conference on Parallel ProcessingJob schedulers play an important role in selecting optimal resources for the submitted jobs. However, most of the current job schedulers do not consider job-specific characteristics such as communication patterns during resource allocation. This often ...
Slurm Simulator: Improving Slurm Scheduler Performance on Large HPC systems by Utilization of Multiple Controllers and Node Sharing
PEARC '18: Proceedings of the Practice and Experience on Advanced Research ComputingA Slurm simulator was used to study the potential benefits of using multiple Slurm controllers and node-sharing on the TACC Stampede 2 system. Splitting a large cluster into smaller sub-clusters with separate Slurm controllers can offer better ...
DP-Fair: a unifying theory for optimal hard real-time multiprocessor scheduling
We consider the problem of optimal real-time scheduling of periodic and sporadic tasks on identical multiprocessors. A number of recent papers have used the notions of fluid scheduling and deadline partitioning to guarantee optimality and improve ...
Comments