GeneralRun jobsJob Submission and Execution
Barbora NG - Job Submission and Execution
Introduction
This page describes Barbora NG cluster’s specific Slurm settings and usage. General information about Slurm usage at IT4Innovations can be found at Slurm Job Submission and Execution.
Node Partitioning
Division of nodes means that if two users allocate a portion of the same node, they can see each other’s running processes. If this solution is inconvenient for you, consider allocating a whole node.
On Barbora NG, SLURM partitions each compute node into smaller units based on NUMA (Non-Uniform Memory Access) domains. Each compute node is divided into 6 partitions, each consisting of 32 CPUs and 122 GB RAM.
Job Submission Options
Jobs can be submitted with different resource requirements:
- No specification: Defaults to one partition (1/6 of a node);
- CPU-based allocation:
--cpus-per-task=<ncpus>or--mincpus=<n>flags allocates the number of partitions needed to satisfy the CPU requirement; - Memory-based allocation:
--mem=<X>flag allocates the number of partitions needed to satisfy the memory requirement; - Exclusive access:
--exclusiveoption reserves the entire node, i.e. all 6 partitions.
