)
Who do you think should win the 2025 Superuser Awards? The annual Superuser Awards are to recognize organizations that have used open infrastructure to improve their business while contributing back to the community.
This year, the Superuser Awards winner will be announced at the OpenInfra Summit Europe, October 17-19! Join us at the annual OpenInfra Summit for an opportunity to collaborate directly with the international community of people building and running open source infrastructure using Linux, StarlingX, OpenStack, Kubernetes, Kata Containers and 30+ other technologies. Get your Summit tickets now!
G-Research is one of the four nominees for the Superuser Awards 2025. Check out why its team is getting nominated:
Who is the nominee?
G-Research: The engineering teams that architect and manage our infrastructure are led by Ross Martyn, with Aaron Scaife, Scott Solkhon, Andrew Ellard, Phillip Quinney, Manuel Perrot and Ian Watson in London, UK, and Ronaldo Lauture, Campbell Krueger, Javier Diaz Jr., Taseen Siddiqui and Sai Oruganti in Dallas, TX, USA. Also, G-Research Open Source Software (GR-OSS), led by Alex Scammon has a development team dedicated to keeping OpenStack maintained and secure, including Jay Faulkner, Afonne-CID, Clif Houck, and Adam MacArthur.
Our success in OpenStack has also been heavily assisted by the team at StackHPC, including John Garbutt, Doug Szumski and Will Szumski.
How has open infrastructure transformed the organization’s business?
G-Research benefits by having greater flexibility, privacy, and the ability to optimize performance – all attributes of an on-site, self-hosted cloud. OpenStack is the obvious choice for such needs. Using OpenStack as part of a larger cloud infrastructure stack allows G-Research to focus on what they do best: using state of the art machine learning to perform quantitative research.
Because of OpenStack’s long history of widespread use, G-Research is able to find and onboard new engineers who already have the experience needed to operate infrastructure at scale. Relying on open infrastructure allows the firm to focus on honing its competitive advantage – the quantitative, machine learning research being performed by industry-leading experts.
How has the organization participated in or contributed to an open source project?
GR-OSS was founded over 5 years ago and dedicated to contributing to open source technologies that power G-Research. These contributions span the open source ecosystem, promoting technical excellence, improving security, and ensuring proper community governance.
Specific examples include:
- Creating Armada, a CNCF Sandbox project
- Innovating in the data science, contributing tp MLflow, FasttrackML, and Polars among others
- Supporting efforts by experts like David Leadbeater to find CVEs in OSS projects, such as git (CVE-2025-48384)
For OpenStack, GR-OSS employs four contributors who build Ironic, help enhance Go SDKs, and focus on necessary security and maintenance work. Over the last 5 years, we’ve contributed over 1,000 changesets and 3,000 reviews to the OpenStack community.
What open source technologies does the organization use in its open infrastructure environment?
G-Research’s technical estate is built entirely on a foundation of open source software. We start with a fully integrated OpenStack, including Ironic, Nova, Neutron, Glance, and Ceph. Many common interactions with OpenStack, such as onboarding new hardware, are made push-button easy by utilizing Jenkins jobs executing a combination of internal and kolla-ansible scripts. Once onboarded, those machines are deployed using Ironic’s direct deploy driver. These deployments happen constantly as G-Research performs a rebuild of every server in the estate on a rolling basis to ensure they have the freshest OS and firmware – via this process, our clusters perform over 25,000 server rebuilds per month.
What is the scale of your open infrastructure environment?
G-Research operates 30MW+ of data centre capacity with 13,000+ GPUs, 450,000 CPU cores, and 2,000 inference accelerators, supported by 240PB+ of storage and 15TB/s throughput. This powers 50M+ HPC/data science jobs daily and Kafka clusters processing 90B+ events per day. All are managed in an OpenStack-based cloud (Ironic, Nova, Glance, Keystone, Ceph), forming the foundation for Kubernetes and Armada, which orchestrate workloads at scale.
What kind of operational challenges have you overcome during your experience with open infrastructure?
Our engineering team developed a tool, HyperSniper, which ensures hypervisors are running supported OS versions. Our Prometheus-based monitoring gathers data about OpenStack via openstack-exporter. This data is used to identify hosts with nearly EOL versions – those are disabled and permitted to drain. After a time, remaining instances are live migrated and the hypervisor is rebuilt and returned to service. This rolling, unattended upgrade process ensures we’re always running a secure OS. G-Research also has, at times, outscaled Kubernetes. To mitigate this, we wrote Armada, a CNCF incubating project for scheduling of batch jobs across multiple Kubernetes clusters. This layered approach with OpenStack, Kubernetes, and Armada provides the flexibility to scale at multiple levels of the stack
How is this team innovating with open infrastructure?
“Innovation requires experimentation,” notes Mark Burnett, G-Research’s Head of Technology and Innovation. Open source enables us to tailor software for unique ML and data science challenges. This includes combining domain expertise with OSS contributions to deliver better solutions for all. GR-OSS itself is also innovative: rather than just consuming OSS, we proactively invest in securing, maintaining, and improving it, sustaining both our ecosystem and the community. Quality software requires consistent effort.
- Meet the 2025 Superuser Awards Nominees - September 16, 2025
- 2025 Superuser Awards Nominee: CIR team at Télécom Paris - September 16, 2025
- 2025 Superuser Awards Nominee: STACKIT IaaS teams - September 16, 2025
