On-premise high-performance computing (HPC) platforms, resilient storage solutions, and cloud computing are inherently expensive. It is why the Research e-Infrastructure strand has seen some of the largest investments in the programme.
Recent communications from RLP have sought to explain the services that the programme has delivered, rather than the individual projects. Here, we have returned to project names to help clearly articulate the journey and improvements from 2018, before RLP started, to today.
Speaking about this workstream Robert Haines, Associate Director of Research IT, said:
‘The Research Lifecycle Programme has made a significant impact on the computing and storage platforms used by researchers across the University. The coordinated investment has enabled us to develop systems that work together and brought significant improvements to our overall capability.’
‘The University’s investment in central platforms like the Computational Shared Facility has enabled us to offer resources free at the point of use for researchers. We have demonstrated that collaboration increases capacity for everyone, and well beyond the value an individual researcher or school contribution.’
Project M&K – Incremental Increase in Computational Capacity
Project M&K was developed to address the continuing increase in demand for computational resources at the University and has had regular investments in each financial year. When the project was originally scoped in 2020 there was an urgent need to increase capacity to address changes in working practices as a result of Covid-19 restrictions, in addition to the normal growth requirements. The scoping exercise also identified significant supplementary benefits:
· Shorter wait times for HPC access
· Increased scale of problems that can be addressed
· Make accessing platforms easier for non-specialists
· Reduce the number of unsupported (Shadow IT) platforms in the University
· Reduce the risk of data breaches
· Increase the success rate of research grant applications
In the first year of RLP the M&K projects funded 16 NVIDIA GPUs, three very high RAM nodes, invested £70k in cloud resources and a further £60k in standard compute nodes. Each subsequent year has seen further investment in these computing platforms totalling £390k on cloud, and procuring a total of 832 new CPU cores, 7 high RAM CPU nodes and 32 NIVIDIA GPUs.
There will be more investments in cloud during the final year of RLP as well as the purchase of additional high RAM CPU nodes.
RLP also helped to bring together funding to develop the University’s Computational Shared Facility (CSF), providing around 15% of the funds to develop the CSF with new compute nodes and NVIDIA graphics processing units (GPUs). This investment helps researchers to tackle larger problems, or smaller problems more quickly, than was possible before. The addition of GPUs also improves the University’s capability to perform research around artificial intelligence and machine learning.
Project N – Edge Compute
One of the significant challenges facing Research IT (RIT) is the incredible diversity of platforms, storage solutions and instruments used across the University. Historically, researchers and faculties have procured bespoke solutions at a local level when the local expertise leaves the organisation, and Research IT are often left to pick up the pieces. RLP’s Project N set out to address this issue by finding ways to procure, support and manage these bespoke solutions through RIT, freeing up researchers to focus on their projects.
Edge computing is a perfect solution to these challenges. Working with researchers, specialists from Research IT design, build and support local compute and storage platforms, customised for, and used exclusively by the group. Edge is particularly useful when particular computation or data-intensive workflows are impractical on central platforms.
Project O – Cloud Compute
Cloud computing is another area where RLP has fostered new innovations for the University. Using Cloud platforms, such as Amazon Web Services (AWS) is recognised as a way of trialling new methods, new technologies, or simply skipping a queue. Often, researchers have used a departmental credit card, to start these tests without realising it is contrary to University policy or how quickly the costs can spiral.
RIT developed a coordinated approach to using cloud across the University. They created an administrative and legal framework that ensures researchers comply with University policies and RIT can support their work, without fear of cost overruns or unexpected charges. Further, researchers receive monthly invoices for their cloud resources that are paid directly from research funds. Regular billing helps them to plan future expenditure and grant costings. (Consider legal compliance)
Project Z High Performance Compute Investments
One of RLP’s largest single investments was made in the University’s HPC pool, which is free at the point of use for researchers. The HPC Pool includes 4096 CPU cores, 192GB of RAM and rapid interconnects and increases the scale of problems that researchers can tackle.
Find out more about the other work streams at: