Computational Physics

Blog

Software

Projects

Courses

Theses

People

About

Events

Internals Getting started Computing Cluster Coding style

Getting started

Computing Cluster

Coding style

This page is a short introduction on how to use Smaug, our in-house computing cluster.

To get access to Smaug, you will need to send us an e-mail with your name, UiO username, phone number, room number and affiliation to the research group. In return, you will receive a password you may use to access the cluster.

The cluster is accessed through SSH. You will first need to access any computer on the UiO network, then any computer in the Physics department, or tid.uio.no. From there, you connect to fyslab-compphys. The following commands should work wherever you are in the world.

Note: If spole.uio.no doesn’t work for you, try tid.uio.no instead.

If this involves too many steps for you, just add the following to your ~/.ssh/config file and change yourusername to your user name:

You should now be able to log in using simply

Once logged in, you should be on the login node, Zoidberg. Do not run your code directly on this node!

First you need to create a job script. Simply create a file with any name, and add the contents you want. One simple example of a job script is this:

This job script will only run on the smaug-a cluster, with 64 processors (that’s all of them), for a maximum time of 1 hour. The job name will be test and it will run the executable “hellow_world_mpi” with MPI. See the files in the folder

on the cluster for a more detailed example of how to write and compile a C++ program that uses MPI.

You submit your job to the queue by typing

if the name of your file is “jobfile”.

If you are expecting your job to finish in less than 6 hours, you may add it directly to the queue. Otherwise, add your job to the calendar . Also please check the calendar before issuing your job.

Slurm has 3 partitions available: normal, smaug-a and smaug-b. Running the job on normal will use all available nodes, while smaug-a and smaug-b will run on a group of 8 nodes, respectively. The default is normal, so unless you really need to run on all the nodes, please specify the partition as smaug-a or smaug-b.

To check the queue for your jobs or others, you simply use the squeue command:

To see only your own jobs, you may specify your username by

A complete list of commands is found here .

You have two options; to run the job on one of the helper nodes (fry, gizmo or leela) or to run it on one or more of the groups (smaug-a, smaug-b, smaug-c or smaug-d).

First, access the node you want to run on. If you don’t need to run it in parallel on multiple machines, you may just start the job as usual.

If you need to run the job on multiple machines in parallel, connect to the first machine in the group, for instance smaug-a1 for the smaug-a group:

From here, start the job with mpirun and provide it with a hostfile:

An example of an hostfile is as follows:

This will tell MPI to run 64 parallell processes, and that it should run on all nodes in the smaug-a group, where each node has 8 processors.

We support installing all packages that are provided in the default Ubuntu repository and are not conflicting with any other installed package. If you’re missing a package, send us an e-mail .

Education Prize to Computational Physics!

Creating a 3D Neuron Scene in Blender

National education price awarded to Computational Physics professor

Choosing the right license for your code

Getting MCALab up and running on a 64-bit installation of Matlab on Ubuntu

Find Usages Keybind broken in QtCreator

PyCharm Keymap for QtCreator Fanatics

Fixing GDB issues with GCC 4.8.x and C++11

LaTeX package for UiO Master’s theses

Circumventing annoying library path issues with QtCreator

Log in

Entries RSS

Comments RSS

WordPress.org