Introduction to High Performance Computing: Glossary

High Performance Computing (HPC) typically involves connecting remotely to a cluster of computers.

HPCs can be used to do work that would either be impossible or much slower in a Desktop environment.

Typical HPC workflows involve submitting “jobs” to a scheduler which queues/priortises the “jobs” of all users.

The standard method of interacting with HPC’s is via a Linux-based Command Line Interface called “The Shell”.

We connect to remote servers, like HPC’s, using the Terminal

SSH is a secure protocl for connecting to remote servers

To connect to a server, you need it’s address, an open port, and your user ID

scp (The Secure Copy Program) is a standard way to securely transfer data to remote HPC systems.

File ownership is an important component of a shared computing space and can be controlled with chgrp and chown.

Scripts are mostly just lists of commands from the command line in the order they are to be performed.

A cluster is a set of networked machines.

Clusters typically provide a login node and a set of worker nodes.

Files saved on one node are available everywhere.

The scheduler handles how compute resources are shared between users.

Everything you do should be run through the scheduler.

A job is just a shell script.

If in doubt, request more resources than you will need.

Load software with module load softwareName

Unload software with module purge

The module system handles software versioning and package conflicts for you automatically.

You can edit your .bashrc file to automatically load a software package.

wget downloads a file from the internet.

sftp/scp transfer files to and from your computer.

You can use an SFTP client like FileZilla to transfer files through a GUI.

The smaller your job, the faster it will schedule.

Don’t run stuff on the login node.

Again, don’t run stuff on the login node.

Don’t be a bad person and run stuff on the login node.

First key point.