Close
The Data Safe Haven Refresh project is delivering several enhanced capabilities and service improvements.
New enhanced capabilities
Infrastructure
The DSH core infrastructure hosting platform now consists of 360 CPU cores and 7.5TB of RAM. These resources are split across 2 data centre locations for a degree of resilience for core services.
Four NVIDIA V100 GPU cards have been attached with a total of 128Gb of dedicated GPU memory. Allowing for dedicated visualisation to Citrix desktops. Or processing compute as required.
A new Citrix virtual desktop platform has been built for the environment updating the deployment technology and standardising the environment components.
High-performance and high-throughput computing
The project has developed and delivered the capability to perform high performance computing (HPC)-style jobs within the Data Safe Haven, while preserving data isolation. This enables data intensive workflows to be performed at scale. Access to the HPC service is currently on a request basis and subject to HPC resource availability.
Developer tools in support of data science workloads
To enable data science workloads, new software tools have been deployed in the refreshed DSH.
Software version control is now available through GitLab, allowing code to be stored centrally, and facilitates collaborative software development. For advanced use cases, automated testing and deployment pipelines can be executed when code changes are submitted to GitLab.
Package management is an important consideration in terms of reproducible research. The project has deployed a product – Artifactory – which enables DSH users to install python, R and conda packages. These packages can be ‘pinned’ to specific versions, ensuring the software stack used to answer a research question is known and can be reproduced if required.
GPU-enabled workloads
The new DSH hardware layer has NVIDIA Graphics Processing Units (GPUs) available. These can be used to accelerate computation of ‘GPU-aware’ workflows and/or applications such as machine learning. Access to the GPU cards is currently on a request basis and subject to availability.
Service hosting
The project has developed the capability to quickly deploy new infrastructure for hosting services on behalf of DSH researchers. Deployments are performed using modern techniques – an Infrastructure as Code (IaC) approach, coupled with configuration management - to ensure consistency while rapidly responding to new requests and changes. This capability will reduce the time to deliver new services for studies.
Media data and video analysis
The new DSH hardware layer has NVIDIA Graphics Processing Units (GPUs) available. These improve video playback and analysis. In addition, the underlying network speed within the refreshed DSH ensures smoother processing of media.
Enabling changes
See our proof of concept case studies for new capabilities.