Stride N1, Arefin A1, Antony J1
1Csiro, , Australia
Biography:
Dr Ahmed Arefin is an HPC Specialist within the HPC Systems Team, Scientific Computing Platforms at CSIRO. He holds a PhD in Computer Science (Data-Parallel Computing and GPU Computing) from the University of Newcastle, Australia. His research interests focus on the application of high-performance computing (HPC) to data mining, graph analytics, and scientific visualisation.
Nathan Stride: Senior HPC Team Member, IMT, CSIRO
Dr. Joseph Antony: Tech Lead, HPC Team, IMT CSIRO.
Abstract:
Edge computing plays an important role in research environments where data must be processed close to where it is generated, particularly for GPU-accelerated and data-intensive workloads. This work presents a practical approach for automatically deploying an Edge Digital Science Platform using Base Command Manager (BCM), Kubernetes, Run:AI, and GitHub Actions workflows.
BCM is used to provision and manage the underlying physical infrastructure, including control and worker nodes. Kubernetes provides the container orchestration layer, while Run:AI enables efficient allocation and scheduling of GPU resources for research workloads. Together, these technologies create a flexible and scalable platform capable of supporting a broad range of scientific applications at the edge.
Automation is driven through GitHub Actions workflows that manage cluster configuration, Kubernetes version upgrades, ArgoCD application deployment, and supporting infrastructure services. This GitOps-based approach ensures deployments are consistent, repeatable, and easy to maintain. Infrastructure and application changes are tracked in version control and can be applied automatically across the cluster.
The result is a streamlined provisioning and management process that reduces manual effort and improves operational reliability. The platform can be deployed rapidly, scaled as requirements evolve, and maintained with minimal overhead. This work demonstrates a practical and reproducible method for building GPU-enabled edge computing environments that support modern digital science workloads.
