What will you learn in Learn to Use HPC Systems and Supercomputers Course
Navigate and access HPC systems and supercomputers: logging in, data transfer, and environment modules
Understand HPC hardware and software stacks: cluster components (login, compute, storage nodes), software modules, and job schedulers (PBS & Slurm)
Write and submit batch jobs with PBS and Slurm: job scripts, queues, interactive jobs, arrays, and job management commands
Develop parallel code using OpenMP, MPI, and GPU programming (CUDA): shared-memory, message-passing, and accelerator models
Program Overview
Module 1: Supercomputers and HPC Clusters
⏳ 40 minutes
Topics: Evolution of supercomputing, cluster vs. supercomputer, benefits of HPC-enabled parallelism
Hands-on: Explore historical supercomputers and compare cluster architectures
Module 2: Components of an HPC System
⏳ 50 minutes
Topics: Login, management, compute, and storage nodes; network interconnects; resource partitioning
Hands-on: Connect to a demo cluster, inspect node roles, and verify system topology
Module 3: HPC Software Stack & Environment Modules
⏳ 50 minutes
Topics: Data transfer tools (scp, rsync), module systems, environment setup, available software lists
Hands-on: Load/unload modules, switch software versions, and run a sample application
Module 4: Job Schedulers – PBS & Slurm
⏳ 1 hour
Topics: Batch vs. interactive jobs, PBS commands (
qsub,qstat,qdel), Slurm basics (sbatch,squeue,scancel)Hands-on: Write and submit batch scripts, monitor job states, and run interactive sessions
Module 5: Parallel Programming with OpenMP
⏳ 1 hour
Topics: OpenMP pragmas, work-sharing constructs (
parallel for,sections),reduction, and performance considerationsHands-on: Parallelize a loop-based computation and measure speedup across threads
Module 6: Message Passing with MPI
⏳ 1 hour
Topics: MPI initialization, point-to-point (
send/recv), collective operations,ping-ponglatency testHands-on: Implement an MPI “hello world,” then build a simple ring-communication test
Module 7: GPU Programming with CUDA
⏳ 1 hour
Topics: GPU architecture, CUDA kernels, memory hierarchy, vector addition example
Hands-on: Write and launch a CUDA kernel for vector addition and profile GPU execution
Module 8: Course Wrap-Up & Best Practices
⏳ 20 minutes
Topics: Job-array workflows, environment reproducibility, resource quotas, and optimizing job scripts
Hands-on: Refine your job scripts for array submissions and add resource directives (time, memory)
Get certificate
Job Outlook
HPC User / Research Computing Specialist: $80,000–$120,000/year — manage and execute large-scale computational campaigns
Parallel Application Developer: $90,000–$140,000/year — optimize scientific codes with MPI/OpenMP and GPU acceleration
Computational Scientist / Data Analyst: $85,000–$130,000/year — leverage supercomputing resources for simulation and data-intensive workloads
Specification: Learn to Use HPC Systems and Supercomputers
|
