Huawei Cambridge chief calls for high performance computing roadmap
Cambridge needs a specific high performance computing (HPC) roadmap to ensure suitable future provision for the city’s power-hungry sectors and research communities – and to strengthen and retain our city’s position as one of the premier hubs for both supercomputing and Artificial Intelligence in the UK and Europe.
Addressing a roundtable which I chaired exploring Cambridge’s advanced computing landscape, Henk Koopmans, CEO of Huawei Technologies R & D UK, made the call to arms.
He urged other corporates in the Cambridge ecosystem to collaborate to make this happen – and highlighted the urgent need for improvements around data issues, in particular data governance and data ownership.
Koopmans said: “It’s a duty for us, as well as the other organisations, to make more facilities available for the rest of the ecosystem here in Cambridge and beyond.
“Years ago, we were hoping that the digital catapult was coming to Cambridge. It didn’t happen. So, here’s a challenge: Why don’t we make one happen ourselves?
“It’s important for a centre like Cambridge to have the capacity to keep doing the research we are doing and to continue attracting the right talent. We’ve now got plenty of HPC-enabled organisations that we didn’t have 15-20 years ago so we need to take the initiative to ensure Cambridge is at the forefront of the next generation of supercomputing.
“And I think the topic for everybody, both at organisational level and individual level, is the one around data ownership. We need to do some serious governance on big data and AI because I don’t think the industry has as yet grappled with that. But legislation is coming up very soon both at EU and UK level.”
This roundtable event was hosted by Business Weekly in conjunction with Kao Data, the UK Innovation Corridor’s specialist data centre, which has been specifically engineered to cater for advanced supercomputing, HPC and intensive AI deployments.
The select roundtable gathering included some of the most influential thought leaders in the Cambridge technology ecosystem including Steven Newhouse, Head of Technical Services, EMBL-EBI; Tim Cutts, Head of Scientific Computing, Wellcome Sanger Institute; Peter Clapham, Team Leader for the Informatics Support Group, Wellcome Sanger Institute; Jon Holgate Head of Infrastructure, University of Cambridge; and Sarah Cunningham, Vice-President, Technology Operations Services Group, Arm.
Other important stakeholders attending were Lee Myall, CEO, Kao Data; and Spencer Lamb, Vice-President, Kao Data.
Addressing the roundtable, Dr Newhouse revealed that Hinxton-based EMBL-EBI – which houses one of the world’s most comprehensive public biological data libraries for the scientific community – currently receives 62 million requests a day coming into the web infrastructure from 24 million unique IP addresses over the course of the year.
Some archives additionally receive close to two petabytes of data a month to be stored long-term as public science record – and the bioinformatics research institute uses over 300 petabytes of raw storage, spread across three data centres, to support its archive and deliver the data.
It is estimated that there are now 40 trillion gigabytes (40 zetabytes) of data in the world, in part due to machines signing on to networks and the Internet.
To put this into perspective, that’s as many digital bits as there are stars in the universe or around 57 times the number of grains of sand on all the beaches in the world.
And with this amount conservatively expected to continue doubling in size every two years – coupled with the imminent arrival of exascale computing (some 1,000 times faster and more powerful than the petascale) – a roadmap, as suggested, is even more urgent if Cambridge is to stay ahead of the curve.
Peter Clapham from the Wellcome Sanger Institute and Sarah Cunningham from Arm agreed.
Dr Clapham told the roundtable: “One of the things we are seeing increasingly with events like COVID-19, is the rate at which data growth is happening: it’s not a gentle curve, it’s giant steps.
“Regardless of whether it’s for NHS Test and Trace, epidemiological research or vaccine investigations, it is essential that we can scale up and manage this data in a coherent fashion across borders in what has now become a worldwide collaboration to tackle diseases.
“Cambridge is at the hub of all this – so having a roadmap is key because then you’re working towards a common set of aims.”
Sarah Cunningham added: “Power is a massive consideration that needs to be focused on and a roadmap to show the power infrastructure that’s going to be built up around Cambridge would allow people to do a lot more planning. Supercomputing needs reliable, abundant sources of power – it’s a foundational building block for further growth.”
Jon Holgate, University of Cambridge, stressed the importance of data management and data governance.
He said: “At the University, we have some HPC with ISO 27000 on one storage, specifically around holding data for the life sciences and so there is a clear roadmap for the next five to seven years about holding that – as well as a very clear location and a very clear governance.
“But that’s very different from a roadmap of user data clearly pushed out to the cloud, or physics and astronomy data.”
Enabling the base infrastructure is critical according to Kao Data’s Spencer Lamb. He said: “The number of organisations in Cambridge utilising HPC and GPU-powered AI – research institutes through to high-tech startups, irrespective of size – is like no-other location in the UK.
“Therefore, power availability and, indeed, connectivity are two important issues for a city with desires to become a ‘smart-city’ of the future.
“From a data centre provider’s perspective, having the ability to develop an advanced, HPC-specialist data centre with access to significant amounts of energy – and, crucially, renewable energy – is critical.
“The lead times for the delivery of power infrastructure are challenging and need to be planned well ahead of the city’s supercomputing demand curve.”
As reported by Business Weekly last month, the UK’s most powerful supercomputer NVIDIA Cambridge-1 will be housed at Kao Data from early this year.
The supercomputer will be used for a range of bleeding edge research ventures harnessing AI – triggering major new initiatives including drug discovery and analysis of NHS datasets.
So an HPC roadmap, initiated by this group, seems to be more than timely. Business Weekly would be interested if you would like to be involved in the next phase of the discussion. Email me with your views: tquested [at] businessweekly.co.uk