High Performance Computing: An Introduction

Prerequisites

The course aims to give an introductory overview of High Performance Computing (HPC) in general, and of the facilities of the High Performance Computing Service (HPCS) in particular.

Practical examples of using the HPCS clusters will be used throughout, although it is hoped that much of the content will have applicability to systems elsewhere.

• All current Cambridge University members (departments and colleges)
• Novice users of HPC and anyone who expects to need to use HPC systems at some stage in their research
• Further details regarding eligibility criteria are available

• Attendance of the Unix: Introduction to the Command Line Interface (Self-paced) or equivalent knowledge
• It is desirable to have shell scripting experience equivalent to Unix: Simple Shell Scripting for Scientists

Number of sessions: 2

• Basic concepts:
  • serial, parallel and high throughput workload
  • a quick look at the modern computer (vector/multithread/multicore/multisocket CPUs; nonuniform memory)
  • interconnecting nodes (ethernet, Infiniband, proprietary)
  • how to put it together (shared memory, distributed memory, ccNUMA, single system image)
  • coprocessors (GPUs and similar things)
  • cluster storage
  • job scheduling
• High Performance Computing Service
  • Darwin - an example of an infiniband CPU cluster
  • Wilkes - an example of a dual-rail infiniband GPU cluster
  • Service Levels (free and non-free usage)
  • Help and further information
  • Connecting
  • SSH login
  • File transfer
  • Remote desktop
  • Tunnelling
• Security
  • Client side
  • Server side
• User environment
  • Compilers
  • Environment modules
  • Filesystems
• Software
  • Free
  • Proprietary
• Job submission
  • Batch scheduler (SLURM)
  • How to submit (HTC, MPI, OpenMP, hybrid)
  • Interactive jobs
  • Array jobs
  • Checkpoint/restart

Presentations and practicals

PuTTY MCS Windows or Linux and HPCS Darwin cluster

• Programming expertise is not required (the focus will be on how to run, rather than on what to run)

• One full day

• Unix: Introduction to the Command Line Interface (Self-paced)
• Unix: Simple Shell Scripting for Scientists

• High Performance Computing
• Scientific Computing
• UIS Systems