Skip to content

Latest commit

 

History

History
133 lines (79 loc) · 3.77 KB

compute_canada.md

File metadata and controls

133 lines (79 loc) · 3.77 KB
Raw

Compute Canada

Intro

What is a compute cluster?

cluster architecture Image from SMART Lab.

Getting an account

Where to get help?

Clusters

Available clusters

General-purpose:

Specialized for parallel jobs:

Connect to a cluster

Pick one general-purpose clusters:

  • graham.computecanada.ca
  • cedar.computecanada.ca

Connect using ssh: ssh <username>@graham.computecanada.ca

You are now connected to a login node, as the prompt shows:

[<username>@gra-login1 ~]$

This is a shared computer, which must be used only for lightweight operations such as file transfers, job submissions, etc Data processing or compilation must be done on /compute nodes/.

Access a compute node, interactively

salloc --time=1:0:0 --ntasks=1 --account=def-glatard

The prompt will tell you the name of the computer that was allocated to you [here, gra796]:

[<username>@gra796 ~]$

Access a compute node, in batch mode

Create a script:

echo '#!/bin/bash' > script.sh ; echo echo Hello >> script.sh

Make it executable:

chmod 7555 ./script.sh

Submit it!

sbatch --time=00:30:00 script.sh

Monitor your job(s):

squeue -u <username>

Your job will write its output in your home folder:

$ cat slurm-4811758.out 
Hello

Transfer your data

The easiest way to transfer your files to/from clusters is through scp, a copy command that works over ssh:

scp file.txt <username>@graham.computecanada.ca:.

Use pre-install software

The module command shows the software packages that are already installed on the cluster and ready to use:

module avail

If you want to use a specific software package, say vtk, just type:

module load vtk

Install your code

You might want to install your own code on the cluster. Here are a few options:

  • Source code: Git clone your repository from GitHub on the cluster.
  • Python code: create a virtualenv and install all the dependencies in it with pip.
  • Linux binaries: copy them directly to the cluster.
  • C/C++: if your code needs to be recompiled, submit an interactive job and compile your code there.
  • Install your code in a Singularity container and run it on the cluster (module load singularity).

Clouds

Compute Canada offers an Infrastructure-as-a-Service (IaaS) cloud, that is, a way to start Virtual Machines with various base images. This is useful if you need to host Web services, or if you need a highly-customized software stack that cannot be deployed easily on a cluster.

Account creation

Once you have a Compute Canada account, fill the form here with your CCDB username and email address. This will give you access to one of the Compute Canada clouds:

  • east.cloud.computecanada.ca, or
  • west.cloud.computecanada.ca

Starting VMs