Alces Flight — HPC 2020 — A Year in Review
While this year wasn’t what we expected some truly amazing things still happened. Here’s our 2020 year in review.
It was the worst of times, it was the worst of times — or so the story of 2020 wants to go. A year that started off with lots of promise cut itself off at the three-month marker when the spread of COVID-19 was officially declared a global pandemic. Suddenly, everyone’s priorities were shifted. For the Alces Flight crew our focus changed, but thanks to our intuitive clients and our ability to adapt with the times some really extraordinary things happened for us in HPC. Here’s our review:
Cloud HPC looked to be more about small projects…
We started off in 2020 working with students and researchers on OpenFlightHPC bootcamps in open source HPC environment creation. These events were designed to challenge the way people think about HPC and creative problem solving. OpenFlightHPC is platform agnostic by design, meaning that you can use the open source toolset to build HPC environments on cloud or hardware. So instead of teaching students how to use a specific machine we instead opened up the idea of students designing an environment to solve their problem. Little did we know that this knowledge would be translated into production use only weeks after our bootcamps kicked off.
… COVID-19 pushes cloud use to the forefront of production work.
Almost immediately after the pandemic news broke out the University of Liverpool, owners of one of the most robust hybrid HPC solutions in the UK, put their service to use. Research team members Joshua Bridge and Yanda Meng, led by Dr. Yalin Zheng, utilised the Barkla HPC Cluster and Cloud Barkla HPC Environment to rapidly work through COVID-19 imagining diagnosis models. Brought together through Alces Flight Center, this 4-month project sped up research efforts that normally would take twice as long and enabled the team to push forward to peer review. The team aims to publish their findings and release public toolsets in 2021. A special thanks to AWS for providing research credits that allowed the team to utilise quad-GPU instances for faster results.
The future of HPC found new ways to learn…
The Alces Flight crew has been promoting the work of the CompBioMed Center of Excellence for over three years. Their aim to bring personalised medicine from science fiction to science fact requires huge levels of collaboration and access to some of the most powerful HPC systems in the UK and Europe. To make their goal possible the team regularly hosts training events to promote the power and potential of computational bioscience. With COVID-19 disrupting their plans the team chose to head to cloud and design a system that would be unfederated in design — this means the environment would not mirror any particular system the team had access to. Instead, the cluster environment would focus on optimisation for education. Designed to meet the objectives of each training event, the nUCLeus HPC environment is aiming to reach those who might otherwise not have access to the technology to progress the field… as well as set-up researchers to work in exascale.
… and facilities are now being built for greater end user engagement.
The Northern Ireland High Performance Computing (NI-HPC) facility, a joint project between Queen’s University Belfast and Ulster University, chose a progressive approach when designing their HPC services. Honing in on six emerging fields of research the team architected an approach which places the researcher at the center of the problem-solving process. Surrounding them are the research software engineers (RSEs) who overlay HPC best practice onto the problems being solved so that work done today can evolve as the fields grow in size and importance.
Digital conferences became a thing…
While we are sad we didn’t get to see much of our community during 2020 what pleasantly surprised us was that the virtual conference format allowed our research to be reviewed by a global audience. We were able to showcase where our open source work with OpenFlightHPC had gotten to at ISC’20, and the teams at University of Liverpool, UCL, and the University of Sheffield had their SC’20 talks attended by over 550+ individuals. Liverpool’s project also got notice at the HPCwire Reader’s Choice Awards and was featured at the AWS HPC Days.
… and so did HPC Cluster competitions.
Computing Insight UK (CIUK) had a 2020 goal to launch their first-ever student HPC cluster competition and they succeeded thanks to several committed partnerships that included the Alces Flight crew. We kicked off their competition in October by building an HPC cluster environment on Microsoft Azure that utilised AMD EPYC2 processors — giving both power and confidence to the four teams who were pioneering the first event and doing so remotely!
The HPC Community found ways to continue to support each other…
We saw the rise of RSE SORSE to keep the community active and engaged, the #hpchuddle turned up on Twitter to keep networking alive for several months of lockdown, and Women in HPC stepping in with remote events and pulled off an SC’20 workshop that included dozens of speakers and hundreds of participants.
… and we want to thank everyone.
This year was definitely different but despite everything some really amazing work got done. We want to send out our deep appreciation for our clients, collaborators, and partners who showed up and put in the work towards keeping the field of HPC moving forward. Look forward to seeing you all again soon in 2021.
