November 26, 2019, by Digital Research Team
High Performance Computing via a Windows interface
In this post, we look at how the University has been testing a Windows interface to its High Performance Computer
As data sets become larger and more complex, there is a greater need for more computing power. The University’s High Performance Computer (HPC) allows researchers to distribute their processing, simulations, and computation across hundreds of specialised computing cores. It is particularly well-suited to tasks that involve big data sets and/or require large amounts of memory. The HPC currently has a Linux-only interface, and Linux is indeed the best way to harness the computer’s processing power.
At last count, the majority of users come from the Faculties of Engineering and Science, representing circa 30% and 65% of total HPC use respectively. One of the aims of the Digital Research Team is to broaden access to the University’s digital facilities. To that end, we have been testing a Windows interface to the HPC with volunteers from the School of Geography and the Nottingham University Business School. The interface looks and feels just like a standard Windows PC and requires no knowledge of Linux.
Using a specially adapted, forty-core node – roughly 10 times the computational power of a typical desktop PC – the testers have been running scripts in R and Stata, and providing feedback on their experience. This has been overwhelmingly positive:
‘I think the use of the HPC through the remote desktop is likely to increase usage among PhD students and other early career researchers developing their coding and understanding of multi-core processing’.
‘I would like to see the Windows HPC expanded as it would definitely benefit my work and I can see many more people, new to coding, understanding and utilising it’.
‘It’s just easy. You don’t need to be a geek: you can’t break it’.
‘Accessing and using the machines was really easy – no problems at all’.
‘My laptop currently limits what I try to do; the Windows VM went way beyond that’.
In addition to the overall positive response to the trial, three user profiles have emerged (though these should not be considered mutually exclusive). Some users have simply appreciated having another Windows machine on which to run their analyses. In effect, this frees up their desktop/laptop for other work. The second type of user profits from the increased memory available on the HPC, which allows data sets to be processed that would be too large for their personal computer. The third type of user benefits from access to multiple computer cores. This means that tasks can be parallelised (i.e. run over multiple processors simultaneously), again reducing the amount of time needed for analyses and simulations. In one instance, an analysis that would normally have taken more than twenty-four hours, was reduced to just four hours on the Windows HPC.
Ben Bedwell from the Digital Research Team helped to co-ordinate the pilot test:
‘Our pilot testers have helped us overcome several technical hurdles along the way, and we’re hugely grateful for their input. For some, this was a real step into the unknown, but everyone I’ve spoken to during the pilot has identified at least one way that this type of service could benefit their research’.
The success of the initial trial means that the University is now looking at developing the option into a service. In this way, the benefits identified above will hopefully be shared among all UoN researchers, and offer a new route into the world of high-performance and cloud computing. This should not only have tangible advantages in terms of research outputs and researcher efficiency, but may also encourage some to take the leap into Linux-based high performance computing, where even greater gains are to be made.
For those interested in learning more about the University’s High Performance Computer, please visit the Digital Research SharePoint site, or contact a Digital Research Specialist. There is a Central Short Course offering an introduction to Linux on the HPC. More experienced users should get in touch with the Digital Research Service, which has high-performance computing experts who can advise on particular needs and projects.