Learn to Use HPC Systems and Supercomputers Course Syllabus

This course provides a comprehensive introduction to High-Performance Computing (HPC) systems and supercomputers, designed for beginners aiming to gain hands-on experience with real-world HPC workflows. You'll learn to navigate clusters, manage software environments, submit jobs using PBS and Slurm, and develop parallel applications with OpenMP, MPI, and CUDA. With approximately 6 hours of focused content, the course combines conceptual understanding with practical demonstrations, guiding you from system access to writing efficient parallel code. While the format is non-interactive, the step-by-step demos ensure clarity and reproducibility, preparing you for roles in computational science and high-performance application development.

Module 1: Supercomputers and HPC Clusters

Estimated time: 0.7 hours

• Evolution of supercomputing
• Differences between clusters and supercomputers
• Benefits of HPC-enabled parallelism
• Historical supercomputers and cluster architectures

Module 2: Components of an HPC System

Estimated time: 0.8 hours

• Login, management, compute, and storage nodes
• Network interconnects in HPC clusters
• Resource partitioning and system topology
• Connecting to a demo cluster and inspecting node roles

Module 3: HPC Software Stack & Environment Modules

Estimated time: 0.8 hours

• Data transfer using scp and rsync
• Using environment modules to manage software
• Loading, unloading, and switching software versions
• Running sample applications in configured environments

Module 4: Job Schedulers – PBS & Slurm

Estimated time: 1.0 hours

• Batch vs. interactive jobs
• PBS commands: qsub, qstat, qdel
• Slurm basics: sbatch, squeue, scancel
• Writing, submitting, and managing job scripts

Module 5: Parallel Programming with OpenMP

Estimated time: 1.0 hours

• OpenMP pragmas and compiler directives
• Work-sharing constructs: parallel for, sections
• Reduction operations in parallel regions
• Measuring speedup across threads

Module 6: Message Passing with MPI

Estimated time: 1.0 hours

• MPI initialization and finalization
• Point-to-point communication: send/recv
• Collective communication operations
• Implementing ring communication and ping-pong latency tests

Module 7: GPU Programming with CUDA

Estimated time: 1.0 hours

• GPU architecture overview
• Writing and launching CUDA kernels
• CUDA memory hierarchy
• Vector addition example and GPU execution profiling

Module 8: Course Wrap-Up & Best Practices

Estimated time: 0.3 hours

• Job array workflows
• Environment reproducibility and module management
• Resource quotas and job script optimization
• Refining scripts with time and memory directives

Prerequisites

• Familiarity with Linux command line
• Basic knowledge of programming in C/C++ or Fortran
• Understanding of fundamental computing concepts (CPU, memory, processes)

What You'll Be Able to Do After

• Navigate and access HPC systems via secure shell and transfer data efficiently
• Identify key components of HPC clusters including compute, storage, and network resources
• Use environment modules to manage software versions and dependencies
• Write, submit, and manage batch jobs using both PBS and Slurm schedulers
• Develop and optimize parallel applications using OpenMP, MPI, and CUDA