̽��ֱ�� of Cambridge - High Performance Computing Service (HPCS)

Cambridge receives £10 million in funding for new AI supercomputer

sc604 — Fri, 27 Apr 2018 13:28:23 +0000

̽��ֱ��new AI supercomputer is a £10 million partnership between the Engineering and Physical Sciences Research Council (EPSRC), the Science and Technology Facilities Council (STFC) and the ̽��ֱ��. Capable of solving the largest scientific and industrial challenges at very high speeds, the supercomputer is supported by Cambridge’s Research Computing Service. ̽��ֱ��aim is to help companies to create real business value from advanced computing infrastructures.

̽��ֱ��supercomputer is part of the UK government’s AI Sector Deal, which involves more than 50 leading technology companies and organisations. ̽��ֱ��deal is worth almost £1 billion, including almost £300 million of private sector investment into AI.

“AI research requires supercomputing capacity capable of processing huge amounts of data at very high speeds,” said Dr Paul Calleja, Director of the ̽��ֱ��’s Research Computing Service. “Cambridge’s supercomputer provides researchers with the fast and affordable supercomputing power they need for AI work.”

In addition to computing power, Calleja and his team will provide training, guidance and support Cambridge researchers, and the wider UK IA industry, to make the most of their data.

“AI projects involving Cambridge researchers are already underway,” said Calleja. “In the life sciences we are working on medical imaging analysis and genomics, and in astronomy, AI is being used as part of the Square Kilometre Array project and research to map exoplanets.”

Cambridge is home to the largest technology cluster in Europe. Over the past decade, start-ups based on AI and machine learning, in Cambridge and elsewhere, have seen explosive growth.

“ ̽��ֱ��UK must be at the forefront of emerging technologies, pushing boundaries and harnessing innovation to change people’s lives for the better,” said Secretary of State for Digital, Culture, Media and Sport Matt Hancock. “Artificial Intelligence is at the centre of our plans to make the UK the best place in the world to start and grow a digital business. We have a great track record and are home to some of the world’s biggest names in AI like Deepmind, Swiftkey and Babylon, but there is so much more we can do. By boosting AI skills and data-driven technologies we will make sure that we continue to build a Britain that is shaping the future.”

Building on the commitment made in the government’s modern Industrial Strategy and its AI Grand Challenge, the AI Sector Deal marks the first phase of a major innovation-focused investment drive in AI which aims to help the UK seize the £232 billion opportunity AI offers the UK economy by 2030 (10% of GDP).

̽��ֱ��deal will help establish the UK as a research hotspot, with measures to ensure the innovators and tech entrepreneurs of tomorrow are based in the UK, with investment in the high-level post-graduate skills needed to capitalise on technology’s huge potential.

It includes money for training for 8,000 specialist computer science teachers, 1,000 government-funded AI PhDs by 2025 and a commitment to develop a prestigious global Turing Fellowship programme to attract and retain the best research talent in AI to the UK.

“Artificial intelligence provides limitless opportunities to develop new, efficient and accessible products and services which transform the way we live and work,” said Business and Energy Secretary Greg Clark. “Today’s new deal with industry will ensure we have the right investment, infrastructure and highly-skilled workforce to establish the UK as a driving force in the development and commercial use of artificial intelligence technologies.”

̽��ֱ��UK’s fastest academic supercomputer, based at the ̽��ֱ�� of Cambridge, will be made available to artificial intelligence (AI) technology companies from across the UK, in support of the government’s industrial strategy.

Cambridge’s supercomputer provides researchers with the fast and affordable supercomputing power they need for AI work.

Paul Calleja

Photo by Pietro Jeng on Unsplash

Mix 2 colours

̽��ֱ��text in this work is licensed under a Creative Commons Attribution 4.0 International License. For image use please see separate credits above.

Yes

Five-dimensional black hole could ‘break’ general relativity

sc604 — Fri, 19 Feb 2016 06:00:00 +0000

Researchers have shown how a bizarrely shaped black hole could cause Einstein’s general theory of relativity, a foundation of modern physics, to break down. However, such an object could only exist in a universe with five or more dimensions.

̽��ֱ��researchers, from the ̽��ֱ�� of Cambridge and Queen Mary ̽��ֱ�� of London, have successfully simulated a black hole shaped like a very thin ring, which gives rise to a series of ‘bulges’ connected by strings that become thinner over time. These strings eventually become so thin that they pinch off into a series of miniature black holes, similar to how a thin stream of water from a tap breaks up into droplets.

Ring-shaped black holes were ‘discovered’ by theoretical physicists in 2002, but this is the first time that their dynamics have been successfully simulated using supercomputers. Should this type of black hole form, it would lead to the appearance of a ‘naked singularity’, which would cause the equations behind general relativity to break down. ̽��ֱ��results are published in the journal Physical Review Letters.

View post on imgur.com

General relativity underpins our current understanding of gravity: everything from the estimation of the age of the stars in the universe, to the GPS signals we rely on to help us navigate, is based on Einstein’s equations. In part, the theory tells us that matter warps its surrounding spacetime, and what we call gravity is the effect of that warp. In the 100 years since it was published, general relativity has passed every test that has been thrown at it, but one of its limitations is the existence of singularities.

A singularity is a point where gravity is so intense that space, time, and the laws of physics, break down. General relativity predicts that singularities exist at the centre of black holes, and that they are surrounded by an event horizon – the ‘point of no return’, where the gravitational pull becomes so strong that escape is impossible, meaning that they cannot be observed from the outside.

“As long as singularities stay hidden behind an event horizon, they do not cause trouble and general relativity holds - the ‘cosmic censorship conjecture’ says that this is always the case,” said study co-author Markus Kunesch, a PhD student at Cambridge’s Department of Applied Mathematics and Theoretical Physics (DAMTP). “As long as the cosmic censorship conjecture is valid, we can safely predict the future outside of black holes. Because ultimately, what we’re trying to do in physics is to predict the future given knowledge about the state of the universe now.”

But what if a singularity existed outside of an event horizon? If it did, not only would it be visible from the outside, but it would represent an object that has collapsed to an infinite density, a state which causes the laws of physics to break down. Theoretical physicists have hypothesised that such a thing, called a naked singularity, might exist in higher dimensions.

“If naked singularities exist, general relativity breaks down,” said co-author Saran Tunyasuvunakool, also a PhD student from DAMTP. “And if general relativity breaks down, it would throw everything upside down, because it would no longer have any predictive power – it could no longer be considered as a standalone theory to explain the universe.”

We think of the universe as existing in three dimensions, plus the fourth dimension of time, which together are referred to as spacetime. But, in branches of theoretical physics such as string theory, the universe could be made up of as many as 11 dimensions. Additional dimensions could be large and expansive, or they could be curled up, tiny, and hard to detect. Since humans can only directly perceive three dimensions, the existence of extra dimensions can only be inferred through very high energy experiments, such as those conducted at the Large Hadron Collider.

Einstein’s theory itself does not state how many dimensions there are in the universe, so theoretical physicists have been studying general relativity in higher dimensions to see if cosmic censorship still holds. ̽��ֱ��discovery of ring-shaped black holes in five dimensions led researchers to hypothesise that they could break up and give rise to a naked singularity.

What the Cambridge researchers, along with their co-author Pau Figueras from Queen Mary ̽��ֱ�� of London, have found is that if the ring is thin enough, it can lead to the formation of naked singularities.

Using the COSMOS supercomputer, the researchers were able to perform a full simulation of Einstein’s complete theory in higher dimensions, allowing them to not only confirm that these ‘black rings’ are unstable, but to also identify their eventual fate. Most of the time, a black ring collapses back into a sphere, so that the singularity would stay contained within the event horizon. Only a very thin black ring becomes sufficiently unstable as to form bulges connected by thinner and thinner strings, eventually breaking off and forming a naked singularity. New simulation techniques and computer code were required to handle these extreme shapes.

“ ̽��ֱ��better we get at simulating Einstein’s theory of gravity in higher dimensions, the easier it will be for us to help with advancing new computational techniques – we’re pushing the limits of what you can do on a computer when it comes to Einstein’s theory,” said Tunyasuvunakool. “But if cosmic censorship doesn’t hold in higher dimensions, then maybe we need to look at what’s so special about a four-dimensional universe that means it does hold.”

̽��ֱ��cosmic censorship conjecture is widely expected to be true in our four-dimensional universe, but should it be disproved, an alternative way of explaining the universe would then need to be identified. One possibility is quantum gravity, which approximates Einstein’s equations far away from a singularity, but also provides a description of new physics close to the singularity.

̽��ֱ��COSMOS supercomputer at the ̽��ֱ�� of Cambridge is part of the Science and Technology Facilities Council (STFC) DiRAC HPC Facility.

Inset image: A video of a very thin black ring starting to break up into droplets. In this process a naked singularity is created and weak cosmic censorship is violated. Credit: Pau Figueras, Markus Kunesch, and Saran Tunyasuvunakool

Reference:
Pau Figueras, Markus Kunesch, and Saran Tunyasuvunakool ‘End Point of Black Ring Instabilities and the Weak Cosmic Censorship Conjecture.’ Physical Review Letters (2016). DOI: 10.1103/PhysRevLett.116.071102

Researchers have successfully simulated how a ring-shaped black hole could cause general relativity to break down: assuming the universe contains at least five dimensions, that is.

As long as singularities stay hidden behind an event horizon, they do not cause trouble and general relativity holds

Markus Kunesch

Marc Van Norden

Star field - 4

̽��ֱ��text in this work is licensed under a Creative Commons Attribution 4.0 International License. For image use please see separate credits above.

Yes

Licence type:

Attribution

Masters of the universe

sc604 — Fri, 19 Jun 2015 12:44:01 +0000

Imagine having to design a completely automated system that could take all of the live video from all of the hundreds of thousands of cameras monitoring London, and automatically dispatch an ambulance any time any person falls and hurts themselves, anywhere in the city, without any human intervention whatsoever. That is the scale of the problem facing the team designing the software and computing behind the world’s largest radio telescope.

When it becomes operational in 2023, the Square Kilometre Array (SKA) will probe the origins, evolution and expansion of our universe; test one of the world’s most famous scientific theories; and perhaps even answer the greatest mystery of all — are we alone?

Construction on the massive international project, which involves and is funded by 11 different countries and 100 organisations, will start in 2018. When complete, it will be able to map the sky in unprecedented detail — 10,000 times faster and 50 times more sensitively than any existing radio telescope — and detect extremely weak extraterrestrial signals, greatly expanding our ability to search for planets capable of supporting life.

̽��ֱ��SKA will be co-located in South Africa and Australia, where radio interference is least and views of our galaxy are best. ̽��ֱ��instrument itself will be made up of thousands of dishes that can operate as one gigantic telescope or multiple smaller telescopes — a phenomenon known as astronomical interferometery, which was developed in Cambridge by Sir Martin Ryle almost 70 years ago.

“ ̽��ֱ��SKA is one of the major big data challenges in science,” explains Professor Paul Alexander, who leads the Science Data Processor (SDP) consortium, which is responsible for designing all of the software and computing for the telescope. In 2013, the ̽��ֱ��’s High Performance Computing Service unveiled ‘Wilkes’ — one of the world’s greenest supercomputers with the computing power of 4,000 desktop machines running at once, and a key test-bed for the development of the SKA computing platform.

During its projected 50-year lifespan, the SKA will carry out several experiments to study the nature of the universe. Cambridge researchers will focus on two of these, the first of which will follow hydrogen through billions of years of cosmic time.

“Hydrogen is the raw material from which everything in the universe developed,” says Alexander. “Everything we can see in the universe and everything that we’re made from started out in the form of hydrogen and a small amount of helium. What we want to do is to figure out how that happened.”

̽��ֱ��second of the two experiments will look at pulsars — spinning neutron stars that emit short, quick pulses of radiation. Since the radiation is emitted at regular intervals, pulsars also turn out to be extremely accurate natural clocks, and can be used to test our understanding of space, time and gravity, as proposed by Einstein in his general theory of relativity.

By tracking a pulsar as it orbits a black hole, the telescope will be able to examine general relativity to its absolute limits. As the pulsar moves around the black hole, the SKA will follow how the clock behaves in the very strong gravitational field.

“General relativity tells us that massive objects like black holes warp the space–time around them, and what we call gravity is the effect of that warp,” says Alexander. “This experiment will enable us to test our theory of gravity with much greater precision than ever before, and perhaps even show that our current theories need to be changed.”

Although the SKA experiments will tell us much more than we currently know about the nature of the universe, they also present a massive computing challenge. At any one time, the amount of data gathered from the telescope will be equivalent to five times the global internet traffic, and the SKA’s software must process that vast stream of data quickly enough to keep up with what the telescope is doing.

Moreover, the software also needs to grow and adapt along with the project. ̽��ֱ��first phase of the SKA will be just 10% of the telescope’s total area. Each time the number of dishes on the ground doubles, the computing load will be increased by more than the square of that, meaning that the computing power required for the completed telescope will be more than 100 times what is required for phase one.

“You can always solve a problem by throwing more and more money and computing power at it,” says Alexander. “We have to make it work sensibly as a single system that is completely automated and capable of learning over time what the best way of getting rid of bad data is. At the moment, scientists tend to look at data but we can’t do that with the SKA, because the volumes are just too large.”

̽��ֱ��challenges faced by the SKA team echo those faced in many different fields, and so Alexander’s group is working closely with industrial partners such as Intel and NVIDIA, as well as with academic and funding partners including the Universities of Manchester and Oxford, and the Science and Technology Facilities Council. ̽��ֱ��big data solutions developed by the SKA partners to solve the challenges faced by a massive radio telescope can then be applied across a range of industries.

One of these challenges is how to process data efficiently and affordably, and convert it into images of the sky. ̽��ֱ��target for the first phase of the project is a 300 ‘petaflop’ computer that uses no more than eight megawatts of power: more than 10 times the performance of the world’s current fastest supercomputer, for the same amount of energy. ‘Flops’ (floating point operations per second) are a standard measure of computing performance, and one petaflop is equivalent to a million billion calculations per second.

“ ̽��ֱ��investment in the software behind the SKA is as much as €50 million,” adds Alexander. “And if our system isn’t able to grow and adapt, we’d be throwing that investment away, which is the same problem as anyone in this area faces. We want the solutions we’re developing for understanding the most massive objects in the universe to be applied to any number of the big data challenges that society will face in the years to come.”

Inset images: Artist's impression of the SKA, which will be made up of thousands of dishes that operate as one gigantic telescope (SKA Organisation); Professor Paul Alexander ( ̽��ֱ�� of Cambridge).

̽��ֱ��‘world’s largest IT project’ — a system with the power of one hundred million home computers — may help to unravel many of the mysteries of our universe: how it began, how it developed and whether humanity is alone in the cosmos.

This experiment will enable us to test our theory of gravity with much greater precision than ever before, and perhaps even show that our current theories need to be changed

Paul Alexander

SKA Organisation

Artist's impression of the SKA, which will be made up of thousands of dishes that operate as one gigantic telescope

̽��ֱ��text in this work is licensed under a Creative Commons Attribution 4.0 International License. For image use please see separate credits above.

Yes

Licence type:

Attribution

Blueprints for supercomputing success

bjb42 — Mon, 01 Nov 2010 12:44:19 +0000

High performance computing (HPC) has become a fundamental enabler of research that requires data handling on a massive scale. Helping this process is the focus of the ̽��ֱ��’s HPC Service (HPCS), which supports over 120 Cambridge research groups through its large-scale, state-of-the-art computational and storage technologies.

Now, following a significant partnership with Dell, the newly launched Dell/Cambridge HPC Solution Centre aims to tackle and resolve HPC challenges identified as important by the research community, and share the solutions worldwide through ‘how-to’ white papers and targeted outreach activities. ‘This is an exciting day for Dell as the Solution Centre marks a significant new collaboration aiming to accelerate discovery,’ commented Dell’s Vice-President, Troy West.

̽��ֱ��new partnership brings the expertise of the HPCS team at the ̽��ֱ�� in providing HPC solutions to the Cambridge research community over the past five years together with the deep computing knowledge of one of the largest information technology companies worldwide.

‘We want to ‘shrink wrap’ this collective knowledge into standardised HPC solutions that other research and non-research communities can use as a template to build their own fit-for-purpose solution, without having to go through the expensive and time-consuming exploration process,’ explains Dr Paul Calleja, Director of HPCS and the Dell/Cambridge HPC Solution Centre. ‘In partnering with Dell, we have launched a new concept in HPC solution development that can deliver best-in-class HPC blueprints back to the HPC community.’

̽��ֱ��first white paper has been released and draws on the Centre’s experiences tackling the increasing need to store up to petabytes of data by setting out a detailed recipe describing how to build a ‘storage brick’. ‘Through providing a range of standardised tested solutions,’ Dr Calleja continues, ‘we can guide users through the complex matrix of processors, networks and storage components to build solutions to real-world situations.’

For more information, please contact Kamila Lembrych (kl341@admin.cam.ac.uk), School HPC Administrator, or visit www.hpc.cam.ac.uk/

̽��ֱ�� ̽��ֱ��’s High Performance Computing Service has formed a partnership with Dell to address key challenges in research computing.

HPCS

Darwin supercomputer

This work is licensed under a Creative Commons Licence. If you use this content on your site please link back to this page.

Yes