For optimal reading, please switch to desktop mode.
Presenting Artemis, a modern cloud-native computing platform
established by the University of Sussex to replace its aging
infrastructure. Built in collaboration with StackHPC, the system
utilises OpenStack and Kubernetes to offer a flexible, software-defined
environment for high-performance research. The architecture
integrates robust and high-performance Ceph storage (even out-performing
the existing Lustre back-end), high-speed networking, and automated
management via Infrastructure as Code to ensure reliability and
scalability. Artemis serves a wide array of academic disciplines,
ranging from artificial intelligence and life sciences to the
humanities, by providing a secure and sustainable foundation for
diverse workloads.
A central feature atop of the IaaS, which is currently in the process
of being fully commissioned, is the Azimuth Cloud Portal. Once
complete, it will further empower researchers through self-service
provisioning of specialised compute platforms and customised virtual
environments.
Modernisation of the HPC Infrastructure
For years, the University of Sussex relied on its Apollo2 cluster
to anchor its HPC capabilities. However, the higher education sector
reached a collective crossroads in the summer of 2024 with the
demise of CentOS 7. With the discontinuation of support for the
industry-standard operating system, the University faced more than
a maintenance hurdle; it faced a strategic ultimatum.
Rather than merely refreshing hardware to service existing technical
debt, the University orchestrated a profound paradigm shift. The
result is "Artemis"—a transition from the rigid, high-touch constraints
of legacy bare-metal systems to a cloud-native future. Artemis is
not just a hardware replacement; it is a fundamental democratisation
of compute power, moving the institution from a traditional siloed
model toward a software-defined research ecosystem.
Democratising Discovery through Self-Service
While OpenStack provides the infrastructure as a service element
providing an interface to Research Computing Infrastructure
specialists, at the heart of Artemis lies the Azimuth Cloud Portal,
a cornerstone of the modernisation effort that effectively decouples
complex infrastructure orchestration from the research process.
Historically, provisioning a high-performance environment was a
manual, administrative-heavy task that often acted as a bottleneck
to scientific momentum. With Azimuth, the University will empower
its researchers to bypass traditional gatekeeping. Researchers will
be able to provision their own bespoke environments—from ephemeral
Slurm clusters with Open OnDemand to Kubernetes, JupyterHub instances,
and Virtual Desktop Infrastructure (VDI) for Linux and Windows—on
demand. This shift to self-service transforms IT departments from
a reactive support function into a proactive provider of high-performance
services, radically accelerating the "time-to-science".
A Polyglot Platform: Bridging Particle Physics and the Humanities
Modern research infrastructure must speak every academic language,
and Artemis was engineered as a "Swiss Army knife" to accommodate
a startlingly diverse range of departmental requirements. The
platform effectively synthesises the needs of traditional high-energy
physics simulations with the burgeoning demands of AI and data
science. The contrast in workloads is a testament to the platform's
versatility. In Life Sciences, Artemis manages the significant "data
gravity" of 5TB CryoEM simulations and provides the GPU-intensive
resources required for AlphaFold protein folding. Meanwhile, the
Humanities Lab utilizes a massive 400TB capacity-optimised HDD Ceph
cluster for text analysis and resilient WordPress hosting. From
supporting the ring-fenced resources and IPv6 requirements of
GridPP’s Monte Carlo simulations to providing the raw GPU power for
Large Language Models (LLMs) in Engineering Informatics, Artemis
proves that a single software-defined platform can serve both the
particle physicist and the digital historian with equal precision.
The Walled Garden: Engineering Sovereignty for Sensitive Science
In an era of increasing data scrutiny, Artemis adopts a "walled
garden" architectural philosophy to ensure absolute data sovereignty.
This is particularly critical for the University’s work in Life
Sciences, where the platform must function as a Trusted Research
Environment (TRE) for processing sensitive MRI data. The security
posture is uncompromising. The OpenStack control plane servers are
entirely isolated with no direct internet access, reachable only
via a hardened admin gateway. Security is further bolstered by a
rigorous encryption strategy: External, Internal, and Back-end TLS
are enabled across all OpenStack APIs, RabbitMQ, and Ceph storage
traffic. Combined with Wazuh for continuous vulnerability detection
and security monitoring, this architecture ensures that flexibility
never compromises compliance, providing a secure sanctuary for the
University's most sensitive intellectual property. “Wazuh is so
simple to then integrate with our institution SEIM and provides
cost saving by filtering alerts/logs locally, reliably and providing
a fantastic automated threat/risk detection platform”.
Software-Defined Resilience: The Infrastructure as Code Philosophy
The technical agility of Artemis is a direct result of its
"Infrastructure as Code" (IaC) foundation. Moving away from the
manual configurations of the past, the Sussex team, in partnership
with StackHPC, manages the entire environment—from low-level BIOS
settings to containerised services, using Ansible and GitHub CI/CD
workflows. This shift is most evident in the replacement of their
legacy workload scheduler with the modern StackHPC Slurm Appliance.
Running Slurm on Rocky Linux, this appliance offers a highly
orchestrated, scalable workload manager that integrates with the
University's Microsoft AD LDAP service. To ensure the platform can
handle the throughput requirements of modern AI, the infrastructure
features a multi-layered network design with high-speed operations
and storage running on 2x100GbE links. By utilising OpenStack Kayobe
and Kolla for containerised service orchestration, the University
has built a system that is not only resilient but inherently agile,
capable of evolving at the speed of the software that defines it.
This provided the University’s first ever staging→production workflow
with real testing on the same hardware without impacting users.
The Expert Voice: A Partnership for Innovation
The successful delivery of Artemis was made possible through a
strategic collaboration with StackHPC. This partnership translated
complex institutional requirements into a high-performance reality,
ensuring the University moved beyond a simple upgrade to a total
modernisation of its research capabilities.
“From initial design to ongoing service improvements, StackHPC have
become an essential partner for Sussex in delivering a modern,
flexible and secure research computing platform to our academic
community, built on a deep understanding and appreciation of our
strategic drivers and operational needs.” Clare Gryce, Associate
Director Digital Research Services at the University of Sussex.
Conclusion: A Research Infrastructure Fit for the Future
Artemis represents a definitive leap forward for the University of
Sussex, turning the "CentOS 7 Cliff" into a launchpad for future
innovation. By embracing cloud-native principles and a software-defined
philosophy, the University has secured its place at the forefront
of global research, ready to support the next generation of
breakthroughs in AI.
Reach out to us via Bluesky,
LinkedIn or directly via
our contact page.