EP

HPC Job Submission Examples

These examples accompany this presentation on Parallel Job Workflows.

Two examples demonstrating different approaches to parallel task execution in SLURM.

Example 1: Job Arrays

Uses SLURM job arrays to run multiple tasks in parallel. Each task computes the sum of integers from 1 to N (N=100,200,300).

Key features:

• Parallel execution via --array
• Task identification through SLURM_ARRAY_TASK_ID
• Independent resource allocation per task

Best for: Parameter sweeps, embarrassingly parallel tasks, multiple independent computations.

Example 2: Sequential Tasks

Runs multiple tasks sequentially within a single SLURM job using a for-loop and srun.

Key features:

• Sequential execution via bash loop
• Task parameter passing through custom environment variable
• Single job allocation

Best for: Learning SLURM basics, tasks with dependencies, resource sharing between tasks.

Prerequisites

• SLURM-based HPC cluster
• Python 3.10+
• Basic understanding of HPC concepts

Directory Structure

.
├── README.md
├── example1/
│   ├── run.sbatch       # Job array submission
│   ├── serial_sum.py    # Computation script
│   └── README.md        # Detailed guide
└── example2/
    ├── run.sbatch       # Sequential submission
    ├── serial_sum.py    # Computation script
    └── README.md        # Detailed guide

Choose Example 1 for parallel computing concepts or Example 2 for basic SLURM job management.