Microsoft Customer Story reproduced with permission.

Fast pace cutting-edge research

The University of Bath wanted to upgrade its infrastructure to provide access to the latest technology, including the newest types of virtual machines as they’re released. It now uses Microsoft Azure high-performance computing to empower its academic community to produce cutting-edge research and enhance course curricula at a faster pace than previously possible.

Storage, monitoring, and management needs are delivered with a suite of Azure services, helping its researchers, teachers, and students benefit from new efficiencies, reduced response time for faster research results, and more opportunities to collaborate with the greater academic community.

Dr Stefano Angioni, Research Computing Manager:

We only had 250 compute nodes on-premises, but Azure offers thousands of nodes, so our researchers and instructors don’t have to wait to spin up their H-series, N-series, and F-series VMs.

Improving research and teaching with additional compute nodes and cloud monitoring

At the University of Bath, cutting-edge teaching, research, and innovation come together to inspire the next generation of creative minds and thought leaders. The university wanted to make sure that its academic community of 23,500 students, teachers, and staff could freely pursue knowledge and ideas unhindered by technology, so it expanded its cloud footprint with Microsoft Azure high-performance computing (HPC). In addition to replacing its on-premises resource capacity, the university’s Research Computing team created a dynamic, user-friendly interface that staff and students use to manage their Azure HPC expenses and user accounts.

Named University of the Year in the Sunday Times Good University Guide 2023, the University of Bath has invested heavily in technology to support its world-class academic research program. The university is no stranger to the world of HPC, having invested in supercomputing and hardware for its on-premises environment over the past 15 years. In the university’s search for the perfect cloud platform to support its digital growth strategy, the Research Computing team was attracted to Azure HPC for running large-scale parallel computations with InfiniBand networking.

Dr Roger Jardine, Director of Research Computing:

Now using Azure HPC, we’ve maintained an InfiniBand network, connected our Slurm scheduler, and carried over ways of working that researchers are familiar with.

In 2022, the University of Bath was deep in the planning stages of its Azure HPC rollout. In less than a year, its Research Computing team moved the university from an intermediate supercomputing system in Azure, known as Janus, to Nimbus, its current Linux-based production environment. The university decommissioned its main on-premises cluster and now runs hundreds of workloads and applications with Azure HPC, including computation-heavy chemistry, engineering, and physics research, domain modeling, and software written by researchers themselves.

University of Bath researchers and teachers successfully migrated to their new Azure HPC environments with no downtime and minimal training. Researchers can now test theories and release their findings faster than before with access to thousands of compute nodes in the cloud versus hundreds on-premises. Teachers also can tailor Azure resources to their course curricula almost on the fly.

Dr Angioni explains:

Azure is like having a very big toolbox - if you have lots of tools, you can build anything with them.

The university estimates that its community uses 200 different types of software, a stark contrast to other industries such as engineering that might only use five or fewer. It’s a testament to the university’s dynamic and diverse research community.

Dr Angioni says:

If you name a software program, we have it, and if we don’t, we know somebody will ask for it soon.

To support this breadth of resources, the University of Bath wanted not only considerable compute capacity but always-on monitoring. The Research Computing team uses Grafana and its Azure Monitor data source plug-in, along with Azure Virtual Machines insights and network insights in Azure Monitor, to visualize and protect its full environment and correlate issues at the infrastructure level. The university plans to provide access to some dashboards for all of its users in the future.

Dr Angioni explains:

Azure Monitor is one of our most important services. We use it to monitor our Azure virtual machines, networking, and almost everything else in the system.

Virtual machines and storage take centre stage

Computational needs vary by individual researcher, and the University of Bath appreciates the flexibility to use a range of Azure virtual machines (VMs). The choice of different underlying Intel, NVIDIA, and AMD processors and configurations means that researchers can work in their own ways. From powerful performance and massive scale to optimizations for specific workloads, researchers can quickly access the VMs they need to solve problems and drive innovation.

Dr Angioni explains:

When a new type of VM like the ND H100 v5 or HBv4-series that’s purpose-built for HPC comes out, we can add it to our environment far faster than with an on-premises resource, and researchers can start using it immediately.

All of the university’s VMs are deployed using Azure CycleCloud. Researchers can easily load the software they need into predefined Azure VM images using EasyBuild and get to work. For its storage needs, the university primarily relies on Azure Disk Storage and is looking to offer archival storage with Azure Blob Storage.

Dr Angioni explains:

Storage in Azure is really important for our users and their research workflows because a lot of research councils have quite strict rules on how long files used for research output are kept, and it can be quite a long time.