Slurm is an open-source resource manager designed for high-performance computing (HPC) use. It provides the following scheduler funcionalities: i) allocation and access of computing resources ii) framework and tools for executing and monitoring parallel or serial work, iii) a job queue for efficient resource management. The Slurm scheduler has a modular client-server architecture and a sophisticated multi factor job priority queue. Several hardware architecture from the biggest HPC systems to small Linux workstations is supported. It is used on many of the Top500 HPC sites worldwide.

In my presentation I will introduce the components of the Slurm scheduler: i) scheduler management with multi-core support and general computing resources, ii) accounting management with a multi-cluster setup. In a multi-cluster configuration the accounting database is capable of controlling all clusters form a central location. I will show how HPC systems at NIIF can be connected and configured in a multi-cluster system.

https://computing.llnl.gov/linux/slurm