Remote compute facilities #

There are a number of computers and clusters available for remote use. This page describes them, how they should be used and how to connect. If you have any questions, please contact us.

Contents #

General guidelines #

Cafe #

Cafe ( is our main SSH host, sometimes called a "login node" in other environments. It runs on a shared Faculty of Science blade in the university data centre.

You can ssh to cafe using your university DS credentials, e.g. ssh

As cafe runs in a fairly constrained virtual machine on a shared faculty server, you must not run heavy-weight compute tasks on cafe as this can severely disrupt access for other staff and students. In particular, this includes machine learning and deep learning workloads. If you're unsure about where to run a certain task, please contact us.

Tunnelling via cafe #

You can use cafe to tunnel to other servers, e.g. if you are running a development web server (such as Flask, or NodeJS) on cpunode1 on port 12888, you can establish a tunnel as follows ...

ssh -L

... which will allow you to access your cpunode1 service from your local machine by going to http://localhost:12888

A common use case is to be able to access a devweb database off-campus (using tools like DBeaver, mysql CLI, psql CLI or from an app being developed on your personal computer). The devweb database pages provide port mapping examples for each database server. The following list shows the ports of common services running in the department (assume all hosts are on the domain):

Service Host Port
MySQL devweb2023 3306
PostgreSQL devweb2023 5432
MongoDB devweb2023 27017
OpenSearch (ElasticSearch) devweb2023 9200

For example, to access MySQL while off campus you could run:

ssh -L

You'd then be able to connect your local MySQL client or app to localhost:3306, which would be tunnelled to devweb2023 via cafe.

Port forwarding on Windows #

While we recommend using the bundled OpenSSH client (which comes with recent versions of Windows 10 and all versions of Windows 11), if you are using an older version of Windows then you may need to use a third party SSH client like putty.

If you're using Putty, you should configure it as follows (as an example of port forwarding for MySQL access):

Now, if you click Open and enter your credentials, both an interactive SSH session (where you can enter commands to run on cafe) and a tunnel to MySQL on devweb will be established. You'll be able to connect to MySQL on devweb by pointing your client at localhost:3306.

When you close the connection (or Putty window), your tunnel will be closed.

Using cafe as a Jump Host #

You can also use cafe as a jump host (also known as a bastion host), for example to connect SSH on cpunode1 going via cafe, you could run:

ssh -J

This would open SSH connection to cpunode1 (automatically tunnelled via cafe).

CPU nodes #

Formerly known as Guacamole Linux Labs

We have a number of computers dedicated to CPU-bound loads and remote GUI access. These are generally more suitable than cafe for heavier workloads, although we do ask you to be considerate of other users who might be using them.

You can access the CPU nodes graphically by visiting Guacamole and choosing Linux Labs. This will connect you to an available CPU node.

Please note that when you disconnect (or if your connection is unstable and drops too many packets), your login session is destroyed. We recommend using the remote graphical login facility only from a stable network.

You can also connect via SSH, to any of the CPU nodes listed below. You'll need to connect via either the campus network, the VPN or by using cafe as a jump host.

CPU node list and specification #

Name CPU RAM Storage
cpunode1 Intel Core i5-12500 16GB 512GB NVMe SSD
cpunode2 Intel Core i5-12500 16GB 512GB NVMe SSD
cpunode3 Intel Core i5-12500 16GB 512GB NVMe SSD
cpunode4 Intel Core i5-9500 32GB 512GB NVMe SSD

GPU nodes #

In most cases, we recommend that undergraduate and taught postgraduate (MSc) students use Wee Archie for any computationally demanding projects (ML, DL, etc). Wee Archie is available to taught students and offers computational power far in excess of what we can offer within the department. Access to Wee Archie must be arranged via your project supervisor (not via CIS IT Support).

Staff and PhD / research students should use ARCHIE WeSt for high performance / computationally demanding work.

In addition to the CPU nodes, we operate a number of GPU nodes - higher performance workstations with GPUs suitable for moderate ML/DL workloads (where other resources like Wee Archie or ARCHIE WeSt are unsuitable). Graphical access is available via Guacamole; if granted access, you will see a new connection labelled GPU Nodes.

GPU node list and specification #

Name CPU RAM Storage GPU
gpunode1 Intel Core i7-10700 128GB 1TB NVMe SSD NVIDIA GeForce RTX 3070
gpunode2 Intel Core i7-10700 128GB 1TB NVMe SSD NVIDIA GeForce RTX 3070

Access to these machines is by arrangement only and is offered to students who require this for their project work. If you think that this facility would be useful, please contact us. Please note that these resources are very limited so access is granted on first-come-first-served basis, in cases where we determine that this is the most appropriate place to run your workload.

Considerate use #

Cafe should be used for login, tunnelling and light tasks (like file management and editing) only.

Our CPU and GPU nodes can be used for more intensive workloads, but please be considerate of other users who may be using these machines at the same time. In general, heavy-weight, experimental and intensive programs:

You can nice a process by using the nice command and ionice command (or renice command if it is already running). For example, to nice a program called myresearch you would run nice +19 ionice -c 3 myresearch

Your home directory resides on an NFS server on a network which is shared across a large number of machines; it is not optimised for high throughput experimental IO. Therefore, programs should not write large amounts of data to your home directory. If there is a need to write large amounts of data during your experiment then please create yourself a directory (using your userid as the name) in /scratch and write your experimental data there. /scratch resides on fast, local SSD disks but is not designed for long-term persistent storage (it's not backed up and is liable to be destroyed during regular server maintenance). As such, you must regularly back-up any non-reproducible data to your home directory.

The department reserves the right to stop your experiment if it is impairing the use of the computer or the network, but we will endeavour to let you know when this happens and suggest alternative computing arrangements where possible.


The university operates a regional supercomputer which is available to academia and industry in the West of Scotland. The machine offers extremely high compute performance and parallelisation and is suitable for the most intensive tasks. You can read more about it at Please note that there is usually a cost for accessing ARCHIE. All ARCHIE access enquiries should go directly to the ARCHIE team; students should do this via their project supervisor.