N8 HPC / N8 CIR
The N8 HPC cluster, Polaris, was decommissioned in July 2018. A replacement machine hosted at Durham was proposed but unfortunately this fell through at the last minute due to VAT-related issues.
N8 HPC will continue but as the N8 Centre of Excellence for Computationally Intensive Research (N8 CIR). To find out more about the new Centre and to read a welcome from their new Director, visit their new website,
N8 CIR is currently investigating a joint Cloud investment with several members of N8 - Manchester has allocated £100k per year for three years to this initiative. N8 CIR will investigate options and how to use Cloud cost-effectively and productively for UoM and other N8 researchers with the first step being to look at Condor (HTC) bursting.
The money that was to be invested in a new N8 machine is still available and will be invested into local (UoM) HPC resource (see below).
Research Lifecycle Programme (RLP)
The Research Lifecycle Programme (RLP) case has been approved and there is now a five year strategic plan of investment of over £15m. There are three streams in the RLP:
- e-Research Infrastructure: a combination and inter-working of digitally-based technology; hardware and software, resources (data, services, digital libraries) communications, and the people and organisational structures needed to support modern, internationally leading collaborative research.
- Research Data Lifecycle: covering the processes and technologies of data creation and deposit, management of active data; publication; data repositories and archives; data catalogues and registries.
- Processes and systems that support the administration of research.
Research IT is most heavily involved in the first two themes and these are our areas of investment. There is at least £2.6 million pounds available over the five years for computational resources of which £971k is already allocated (see below).
What are we going to spend it on?
As soon as possible there will be a £971k investment in (true) HPC resources, using the funds previously allocated to N8 HPC. We expect to order in November with the resources likely to be in production in late February. As this will be free at the point of use (FATPOU) the likely access route will be through a light touch application process similar to that previously used for N8 HPC.
Subject to the detailed business cases mentioned above, a further £360k will be invested in computational resources this financial year with a production target of April / May 2019.
The recently held workshops (see below) will determine exactly what kinds of resources these funds are used to procure (e.g. GPUs and high-memory nodes).
The remainder of the RLP funds available for computational resources will be spent over the next three years.
RLP Infrastructure Workshops
Workshops were held in October to obtain a list of priority spending areas for the £360k mentioned above. We are grateful to everyone who came along to these meetings and provided their views.
The main requests were for:
- (GP)GPUs
- High-memory compute nodes, with a minimum of 1 TB RAM.
- Interactive computational resource, including:
What happens now?
On-going Work
Work is currently ongoing to merge the CSF2 and the DPSF, to create CSF3 (which already hosts all compute resources procured this calendar year). The merger of the DPSF is now complete. There will be further details about this on our blog and the RIT website in due course.
Upcoming — Summary
1. £971k invested in HPC resources, expected in production in late February.
2. January procurement of £360k GPU, high-memory and Cloud-based resources — subject to successful business case — expected in production late April.
We hope you have found this summary useful but if you have any further questions then please let us know.