New platform for infrastructure modelling now open

Professor Jim Hall

By Professor Jim Hall, Chair of DAFNI Governance Board

A new tool, set to transform research, planning and policymaking around infrastructure in the UK and beyond, is now in alpha testing.

DAFNI 1.0 from the Data and Analytics Facility for National Infrastructure (DAFNI) takes a significant step forward this autumn as its platform moves into alpha testing with users in eight UK universities. The platform is designed to allow a more cohesive picture of the UK’s infrastructure needs, possibilities and policies to be developed.

The result of over £1m of hardware investment and two years’ development, DAFNI 1.0 is designed to create a cohesive picture for infrastructure needs, possibilities and policies. It is already advancing research and allowing users to run models more quickly, through greater compute capacity and at greater scale and level of detail than has previously been possible.

DAFNI 1.0 has already improved the time it takes to run complex models, meaning it can be used for short-term decision-making as well as long-term planning, saving valuable researcher and computing time.

Much more than a team of developers and massive compute power, DAFNI 1.0 is underpinned by groups across industry, government and academia, and is actively seeking further collaborations. The support of the community that conducts infrastructure research and their active participation in DAFNI 1.0’s dynamic evolution ensures models driven by academia help validate and inform policy and new infrastructure developments.

We’re already working with organisations including the National Infrastructure Commission, the Alan Turing Institute, the Department for Transport, Oxfordshire County Council, the Infrastructure Transitions Research Consortium (ITRC), the University of Oxford, Southampton University and Leeds University.

A number of pilots have run on the DAFNI platform across the last year, with the models now live, including:

• The Digital Communications Model – The implications of roll-out of 5G mobile networks across the UK
• The Agent-based Housing Model – How different factors affect the UK’s housing prices and market forces
• Automated Demand-forecasting Model for New Local Railway Stations – Forecasting demand and identifying the best locations for new local railway stations
• Population Estimation and Scenario Projection Model – Mapping the UK population in terms of future growth, migration and household structures

Transport Scotland is currently evaluating DAFNI’s demand-forecasting model with a view to using it to rapidly generate forecasts of passenger trips at potential new stations under a range of future scenarios; thereby informing decisions on infrastructure requirements, efficient build and spend and most appropriate station location.

The model in its entirety can be linked to other models in DAFNI 1.0, and DAFNI 1.0’s user-friendly interface means that researchers can take a view on interconnected infrastructure networks and derive an integrated vision on infrastructure provision.

DAFNI consists of five key parts:

1. National Infrastructure Database (NID) This is an increasingly large library of datasets (already over 600) – some of which DAFNI holds, some of which DAFNI facilitates access to. Metadata tags are incorporated to make searching the database easier.
2. National Infrastructure Modelling Service (NIMS) This is split into two main parts – the national modelling catalogue; and the workflow section, where you can drag and drop different models to link with each other to provide a “system-of-systems” approach to modelling and analysis. The aim is to democratise access to models whilst allowing modellers to set permissions on who can use their work.
3. National Infrastructure Cloud Environment (NICE) Users can make use of the high-processing computing power behind DAFNI 1.0 to develop and increase the speed of their models, which may involve hundreds of gigabytes and much more. DAFNI 1.0 uses best practices in industry, such as hybrid cloud and onsite investment.
4. National Infrastructure Visualisation Suite (NIVS) Modellers can use GPUs and CPUs to create 3D renders and 2D plots to tell effective and meaningful stories through visualisations.
5. Data Security Service (DSS) This provides security assurances to those providing the data. Data is looked after in a secure and controlled fashion, based on industry best practices.

As one of our collaborators, Nick Cook, Senior Analyst at Tessella, explains, “DAFNI 1.0 lowers the barrier to entry for research, as users don’t need to procure expensive IT infrastructure, set up their own cloud environment or store data. It enshrines provenance and traceability and allows for reuse of workflows and models for wider community to engage with.”

Key facts about DAFNI 1.0

• 50,000 lines of code already in the platform
• Platform runs on Science and Technology Facilities Council (STFC) compute power, which hosts around 40 petabytes of data
• 20 programming languages used
• 20 independent services combining to create the platform
• At the time of writing this article has 600+ datasets and the list is growing steadily

The alpha launch swiftly follows the DAFNI Conference, which was held at the Royal Society in London in June 2019 with an audience of around 200 delegates from government, industry and academia. Watch the video here: www.youtube.com/watch?v=9ZA1UAcGZCw .

Web: www.dafni.ac.uk Email: info@dafni.ac.uk