Research Platforms

Welcome to the University of Leeds’ Research Computing Resource Guide! IT Services provide a number of tools and services for both data storage and compute, Research Computing being the primary point of contact for research users.

Finding the Right Computing Resources for Your Research

Whether you’re working on small-scale analysis, developing complex simulations, or running intensive computations, we have a wide range of computing resources tailored to support your research.

This guide will help you navigate through various options, from individual devices to advanced High-Performance Computing (HPC) systems. To find the most appropriate resource, consider these questions:

Are you looking for something you can use immediately on your personal device?
Do you need more power without the complexity of HPC systems?
Are you running large-scale computations that require multi-node parallel processing?

Use our Resource Roadmap below to discover the best solution for your research needs.

Resource Roadmap

Starting Small: Individual Device
- If you’re just getting started or your computations are lightweight, you can easily begin on your laptop or desktop.
- Ideal for initial code development, small data analysis, and non-intensive simulations.
Need More Power? Try Shared Devices
- If your work is starting to slow down your individual device, consider using a cluster machine or Linux server.
- These shared systems give you extra memory, CPU power, or GPU access without the complexity of an HPC queue. Great for testing and debugging.
Cloud Solutions for Flexibility and Collaboration
- For collaborative projects, cloud resources offer flexible environments.
- These are perfect for sharing work, developing in teams, and quickly testing your code without investing in hardware.
- Unfortunately, most of the solutions are not supported by the university
Need to Scale Up? Local HPC is the Next Step
- When your research requires heavy computation, high-end GPUs, or parallel processing, it’s time to consider our local (Tier 3) HPC resources.
- Get direct support from the Research Computing team and access systems tailored to your project’s needs.
For the Largest Projects: External HPC
- For large, multi-node, or specialized GPU jobs, Tier-2 (Regional) and Tier-1 (National) HPC facilities offer unmatched power and scale.
- These external HPC systems are ideal for massive datasets, complex simulations, or GPU-heavy research.

Detailed Resource Overview

Following the roadmap, we dive into each section, helping you understand the specifics of each option, what benefits they bring, and how they can fit your project’s needs.

Individual Devices

Platform	Application	Information
Standard Laptops	Standard request via University IT - fast delivery/setup Flexible and Portable Impacts your other work whilst running computationally intensive jobs	Request Form
Standard Windows Desktops	Standard request via University IT - fast delivery/setup Impacts your other work whilst running computationally intensive jobs Not necessarily considerably faster than laptops	Request Form
Non-Standard Computer	Request via University IT Custom solution Long Purchasing/setting up time Impacts your other work whilst running computationally intensive jobs	Request Form
Windows Virtual Desktop	Useful for small admin tasks Easy access to online Windows 10 desktop Access to Network drives (e.g. M:, N:), Microsoft Office; on-campus web services, Banner, Argos, Printing, and teaching applications.	Overview

Shared Devices

Platform	Application	Information
Clusters	Useful for trialling and debugging on a high end machine without needing to buy a workstation or queue on the HPC for testing Useful for access to Windows for non-Windows users Access to Printer / Copier / Scanner Possibility of 24h access Access to a range of software (installed or available via AppsAnywhere)	Overview Software List Cluster Map
Linux Workstations/Servers	Benefits of higher specification RAM/cores/GPU; generally high end compared to personal systems None of the queuing issues or other complexity of HPC	Environment System

Cloud Resources

Platform	Application	Features	Information
Google Colab	Jupyter Notebook service useful for teaching, training and development	Can be useful for trialling and experimenting with GPU Python codes Relatively limited CPU (2 vCPU, 13G RAM) No control about resource allocation - bad for reproducibility	Google Colab Homepage
MyBinder	Easier to install dependencies than Colab	Opens up the whole repository rather than just a single notebook Low machine specification as it is 1-2Gbytes of RAM	MyBinder Homepage
GitHub codespaces	Useful for Python, non-Python, non-GPU projects and teaching.	You can easily have 4 cores and 32Gbytes of RAM Provide an environment to assist others to reproduce your work Easy to shift from devcontainers to apptainer containers for use on HPC and Linux systems on campus The University of Leeds has a GitHub Enterprise educational licence	Codespace Overview Codespace Documentation IT GH Enterprise Article
GitHub actions	A model for well defined workflows. CI/CD lifecycle	Automating code testing and deployment Run actions in Windows, Linux and macOS environments The University of Leeds has a GitHub Enterprise educational licence	GH Actions Overview GH Actions Documentation IT GH Enterprise Article

Local HPC

Platform	Application	Features	Information
ARC3/4†	Higher RAM/core/storage than can be achieved in previous options	GPUs (Nvidia Tesla V100), but a relatively small number ARC3 GPUs often lightly used, enabling development and debugging work with interactive sessions Support - ability to get guidance and advice from the Research Computing team on how to best solve your research problems on the platform	Request an Account ARC Documentation
Aire‡	11 times more FLOPS than ARC3 and ARC4 combined	84 Nvidia L40S GPUs Total of 9072 Cores Individual nodes are larger than ARC4 (768GB and 168 cores per node) 2 Large memory nodes (2.3TB)	-

† ARC3 and ARC4 are being decommissioned and the service will be terminated late 2024

‡ Aire will replace ARC3 and ARC4 and it will be available late 2024

External HPC

Platform	Application	Features	Information
Bede	Running multiple GPU jobs, especially across multiple nodes	100% GPU based, with Power9 CPUs adding complexity to building research environments Also contains Grace-Hopper ARM GPU nodes, which are likely the fastest most capable systems you have access to. Login nodes have GPUs, so useful for debugging and development	Bede Overview
JADE2†	Running multiple GPU jobs - simpler than Bede due to the x86\_64 architecture	100% GPU, Nvidia Tesla V100 Higher performance storage options	JADE2 Overview
ARCHER2	Much larger system for parallel CPU jobs	Individual nodes are larger (up to 512G and 128 cores per node) Total of 750k cores	ARCHER2 Overview
Jasmin	A much more comprehensive service for NERC users	Jupyter notebook support HPC Shared Linux systems Cloud	Jasmin Overview

† The JADE2 account registration period closed on 01/09/2024. As of this date, we are no longer able to approve project requests. 1st November 2024: Batch and interactive access to all compute resources will be withdrawn. 6th January 2025: All access to the service will be withdrawn and physical decommissioning of the system will commence.