newsIssue 81 Summer 2017

Centre of Excellence to promote HPC in biomedicine

Building a computational human

The HPC-Europa visitor programme returns!

The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

In this issueEPCC’s role in the international exascale effort

CodeOpt Scotland: powering Scottish industry

Increasing earthquake resilience

2

Contact us

3

7

4

9

8

5

10

14

6

Contents

www.epcc.ed.ac.uk info@epcc.ed.ac.uk +44 (0)131 650 5030 EPCC is a supercomputing centre based at The University of Edinburgh, which is a charitable body registered in Scotland with registration number SC005336.

Director’s WelcomeWelcome to the Summer 2017 edition of EPCC News. I hope you enjoy it.

16

18

20

New staff at EPCCMeet our new colleagues

CodeOpt ScotlandImproving vital business software

Fortissimo MarketplaceCloud-based HPC for industry

HPC-Europa returnsEU visitor programme is back

ARCHER eCSEDeveloping ARCHER software

Exascale and EPCCOur work in this exciting field

BioExcelBuilding a Centre of Excellence in Computational Biomolecular Research

CompBioMedCentre of Excellence in Computational Biomedicine

Seismology appReducing earthquake risks

Next-Generation Sound Synthesis projectAccelerating synthetic music codes with GPUs

Intel Parallel Computing CentreImproving the performance of scientific applications on Intel hardware

New Tier-2 systemCirrus upgraded and transformed

Masters programmesOur two courses are designed to boost careers

Supercomputing MOOCWhat we learned from the first run of our new course

Women In HPC at ISCIncreasing diversity in HPC

EPCC at ISC 2017If you’re attending, say hello

Raspberry Pi clustersHow will you build yours?

HPC outreach reviewOur team has been busy!

MScs at EPCCOur well-established courses lead the field

Come and find us at ISC in Booth A-1363!

Read more on the EPCC blog: https://www.epcc.ed.ac.uk/blog/2017/05/24/scc-isc17

12

17

22

23

24

27

28

It’s a very exciting time for EPCC as we prepare, around Easter 2018, to move to our new purpose-built offices in central Edinburgh at the University’s new “Bayes Centre for Data Science and Technology”.

In this issue, we have stories on a wide range of our projects funded from a variety of sources. I’m particularly proud of the installation of our new EPSRC-funded Tier-2 HPC system, Cirrus, which opened on 1st April 2017 on time and within budget. This funding has extended our SGI ICE XA system, originally installed in April 2016, to 10,080 Intel Broadwell cores.

You will also see that, despite BREXIT, we are still being successful in Europe. The funding of HPC-Europa3 represents the return of our HPC Transnational Access programme which has existed since 1993. We also have a number of very active Exascale research activities – all funded via the European Commission’s Horizon 2020 programme.

I sometimes wonder, as the UK prepares to leave the European Union, if the Government properly understands the scale of some of these projects. Yes, the money is

often spread across multiple partners in Europe but it’s the scale of the R&D that is enabled that is important. The Fortissimo projects have in total 215 partners, engaged in 92 sub-projects with combined funding of €24 million. Although the NEXTGenIO project has only eight partners, it has funding of €8 million. Key in both cases is scale – NEXTGenIO is developing new HPC server technology which is an expensive business, while Fortissimo is focussed on delivering jobs and growth by stimulating adoption of HPC by its partners.

It’s vitally important if the UK wishes to remain globally competitive that Government understands the scale at which it needs to fund projects to replace the lost Horizon 2020 funding. Without this the UK’s innovation potential and the economic growth it’s supposed to drive will be significantly diminished. In truth, I very much hope a way is found for the UK to remain within Horizon 2020 and subsequent programmes.

Prof Mark ParsonsDirector, EPCCm.parsons@epcc.ed.ac.uk

ISC Student Cluster Competition: Go Team EPCC!

3The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

Colin PaulBusiness Development Manager

My role at EPCC is to work closely with industry to gain broader adoption of high performance computing (HPC) and Big Data analytics to solve complex problems for small and medium sized companies.

Prior to EPCC I spent 20 years working internationally in senior sales and application development roles for a variety of fascinating software start-ups.

Outside of work I love travelling, cycling and photography.

Ben MorseMSc Programmes Officer

I joined EPCC from the Royal (Dick) School of Veterinary Studies where I supported staff and students in the fourth year of the undergraduate programme. I now support EPCC’s MSc Programmes in High Performance Computing (HPC) and HPC with Data Science, including marketing, student support, and administration.

My interests include sports (American football, rugby, football), gaming, and music (I sing with a local choir, and two local Gilbert and Sullivan groups).

Laura MainAdministrative Assistant

I have been in the post of Administrative Assistant since February and thoroughly enjoy it. I am working in the Admin office and in Accounts. Before starting here, I worked for the Energy Saving Trust for four years, starting off in the admin team then moving to an office coordinator role.

Outside of work I enjoy socialising with friends. I’m also a big foodie and I like trying out new places to eat.

Emilio PerezLinux Systems Administrator

I am an enthusiastic Linux Systems administrator working at EPCC’s Advanced Computing Facility. I enjoy setting up containers and networks, researching new HPC technologies, and helping support and maintain our infrastructure. Before EPCC I worked in the Apps Division of the University of Edinburgh.

When I am not at work I enjoy activities such as trekking, cycling and travelling around the Highlands, cinema, concerts and computer games.

Angela BorlandFinance Officer

I joined the Accounts Team in December 2016. I perform all sorts of accounting tasks and I am involved in the finances of our various projects. So far it has all been interesting and enjoyable! And busy! Before joining EPCC I worked in the University’s Tax and Payments departments, and before that in various financial and management accounting roles.

My interests out of work are all types of dancing with the exception of Morris.

William KerrSystems Developer

I recently started working for the Systems Development Team as a Systems Developer. Before that, I was part of the Computing Support Team in the School of Physics & Astronomy at the University of Edinburgh. My first few weeks have provided new challenges but change is a good thing as they say.

My personal interests include politics, history & new technology.

Introducing...Meet our six newest colleagues.

4

George Graham Commercial Manager, EPCCg.graham@epcc.ed.ac.uk

CodeOpt Scotland is a partnership between Scottish Enterprise (SE), EPCC and Intel. As a collaborative programme, it aims to generate significant benefits for all partners.

Benefits for companies

For companies, optimised software can potentially reduce costs, increase revenues, improve products and services, or shorten time to market. The key success factor is to select business critical software that is directly related to company success, regardless of whether the code is utilised in the R&D function, in operations, or in sales and marketing.

Close collaboration within the optimisation project is also an important factor, with the company’s own software experts working hand-in-hand with EPCC and Intel experts. SE, EPCC and Intel contribute to the project funding, and ideally the company provides the in-kind contribution of its in-house experts.

Optimising processor design

For Intel, investigating how key codes perform on processing platforms can generate significant insights into how to develop new,

better performing, processors.

It is no coincidence that Intel has chosen Scotland as the focus for this intelligence gathering exercise. Intel is optimising platforms to cater for tomorrow’s advanced data analytics and machine learning applications, and Scotland is already home to many of the leading firms in the financial services, energy, oil, and gas sectors, all of whom are generating significant advantages from such technologies.

Benefits for Scotland and EPCC

An increase in company competitiveness will have a direct effect on the Scottish economy through jobs growth and increased output. Attracting global leaders such as Intel to Scotland will reinforce the strength of the Scottish digital technology sector in which numerous start-ups in the digital, data and novel computation space are already emerging. For EPCC, the opportunity to be at the centre of a resurgent Scottish digital technology sector will further reinforce our position as one of Europe’s most successful high performance computing and advanced data analytics centres.

The CodeOpt Scotland programme is designed to increase company competitiveness by optimising the performance of business-critical software. Launched in March, it is already proving a hit with Scotland’s leading companies.

Image: istock.com/Antiv3D

CodeOpt Scotland: powering Scottish

industry

We have been helping local, national and international businesses for more than 25 years.

Find out more about our services for industry on our website:www.epcc.ed.ac.uk

5The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

The Marketplace was developed during the Fortissimo project, which ended in December 2016. It offers advanced modelling and simulation services to industry, with an initial focus on manufacturing applications.

Users of the Marketplace pay only for the amount of service they need. This is a potential game-changer for many companies, as it allows them to trial the use of advanced simulation and modelling with very low risks, and allows them to build up their use steadily.

Spreading the benefits of HPC

The Fortissimo Marketplace gives access to services in a way that avoids high-risk investments and disruptive step-changes in the way companies work.

The result will be that the benefits of HPC are opened up to small and medium sized companies that cannot afford to run their own HPC systems. This support for small companies, which are recognised as the driving force behind economic growth, is one of the key objectives of the Innovation for Manufacturing SMEs (I4MS) programme which funded Fortissimo.

The Fortissimo project carried out

over 50 experiments to prove the concept of providing services on cloud-based HPC, with many of these experiments leading to commercial services that are being brought into the Marketplace.

As reported in previous issues of EPCC News, the results of Fortissimo show that end-users can gain rapid benefits including:

• reduced time to market for new products

• improved product quality

• lower costs of production

• more efficient use of materials.

Open for business

The Marketplace is open to new suppliers as well as end-users. Software owners and providers can develop their own solutions based on the HPC systems that power Fortissimo, meaning they can scale up to meet the demand for their services without having to invest in expensive new equipment. Instead they can focus on their application expertise and provide new innovative services that will help expand their customer base.

For users and suppliers of HPC application services, the Fortissimo Marketplace is now the place to do business.

Mark SawyerBusiness Development Manager & Project Manager, EPCCm.sawyer@epcc.ed.ac.uk

Fortissimo Marketplace: Cloud-based HPC for business

www.fortissimo-marketplace.com

www.i4ms.eu

Image courtesy Koenigsegg.

Europe’s major HPC centres have joined with software vendors and application experts to create the Fortissimo Marketplace, which offers cloud-based HPC services to industry.

Simulation image courtesy Automobili Lamborghini.

Fortissimo is a Framework 7 project. It has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement No 609029.

6

Many readers will remember the HPC-Europa programme and its predecessor TRACS, which together supported more than 850 research visits to EPCC between 1993 and 2012.

Over the last few years HPC-Europa has been much missed, judging by the number of former visitors and their colleagues and local host research groups who have contacted us to ask if there will be a new programme.

This clear evidence of a continuing demand even after a gap of several years confirms what we have always believed: HPC-Europa offers a unique and highly valuable opportunity for both visiting and hosting researchers.

HPC-Europa3 again allows researchers to spend up to three months in a participating country to collaborate with others working in a similar field, while making use of the project partners’ world-class HPC facilities.

EPCC alone will support 200 research visits over four years, while UK-based researchers will be able to apply for visits to associated institutes in Finland, Germany, Italy,

the Netherlands and Spain.

The programme is open to researchers of all levels, from any discipline which can make use of HPC. Travel expenses and a living allowance are paid.

In HPC-Europa3, EPCC is joined by ICHEC, the Irish Centre for High-End Computing, to form a “British Isles” region. This increases the range of available computing facilities, as well as introducing a whole new network of potential host research groups in Ireland.

Regional Access Programme component

To encourage the uptake of HPC in under-represented countries, a new Regional Access Programme component focuses on the Western Balkans and Baltic regions (represented by GRNET in Greece, and KTH-PDC in Sweden respectively). This offers smaller allocations of HPC resources to users within those regions who have less experience, with a lower threshold for participation. Experienced users in those regions will still be able to apply to the regular HPC-Europa programme.

We are delighted to announce the return of the much-loved HPC-Europa Transnational Access visitor programme.

HPC-Europa returns!

Catherine InglisProject Manager, EPCCc.inglis@epcc.ed.ac.uk

To find out more, see the programme website:

www.hpc-europa.org

HPC-Europa3 started on 1st May 2017. The first closing date for applications is expected to be in early September, with the first visits beginning in late 2017.

HPCInfrastructure on High Performance Computing

Europa

Pan

-Eu

rop

ean

Res

earc

h

7The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

Typical eCSE projects improve the performance, usability or long-term sustainability of software that runs on ARCHER, or add new science-enabling functionality to it. Communities new to ARCHER can apply for funding to, for example, move software from local compute facilities to ARCHER.

In each funding call applicants are invited to submit proposals to a panel of experts drawn from high-performance computing (HPC) and a wide range of computational science domains. Here at EPCC, the ARCHER eCSE Team oversees the entire programme, from opening calls, organising application reviews and panel meetings, to setting up and closing the projects of successful proposals, while providing feedback and advice to applicants. Throughout the lifetime of a project, we liaise with technical development staff and run many useful courses and webinars. When projects end, a final report is produced and a webinar summarising the project is given.

AchievementsMany of the 10 most used application codes on ARCHER have been the focus of an eCSE project.

Software with more modest user bases have improved user uptake and widened their impact through eCSE-funded work. Furthermore, performance improvements can lead to tens of thousands of pounds of savings in compute time.

New HPC communitiesThe benefits and impact of eCSE-funded work are felt beyond the ARCHER programme, in the wider academic and industrial communities and even in traditionally non-HPC communities.

For example, EPCC and the University of Hull moved VOX-FE, a Voxel-based Finite Element bone modelling suite, from a local desktop to ARCHER, dramatically improving its performance and functionality. VOX-FE can now analyse very large, high resolution models with accurate geometry. This, together with its new adaptive remodelling functionality, could make VOX-FE a novel way for paleobiologists to carry out in silico reconstruction experiments of partially recovered bone from dinosaurs and other fossils.

1. Prof. Michael J. Fagan’s Medical and Biological Engineering Group, University of Hull.

Dragonfly head capsule (width 2 mm) with 3D stress and strain patterns modelled by VOX-FE

on ARCHER and visualised with VOX-FE’s ParaView-based GUI. This work was funded by the eCSE programme and carried out by

Neelofer Banglawala and Iain Bethune (EPCC) and Prof. Michael J. Fagan and Richard

Holbrey (University of Hull).

Several times a year, UK researchers can apply to the ARCHER Embedded Computational Science and Engineering (eCSE) programme which funds the development and improvement of scientific software running on ARCHER, the UK’s national HPC service.

Better software, better science

eCSE CallsFor highlights of past projects and information about future calls see:

www.archer.ac.uk/community/eCSE

As I write, the programme is on its 11th call. Over the past 10 calls, we have processed 160 proposals, of which 78 were successfully funded, involving 25 UK institutions. A total of 797 person months of effort were awarded to develop over 40 different scientific application codes, covering molecular dynamics simulations of biological processes, blood flow, tsunami and coastal modelling, first-principles simulation of materials, climate modelling, biomechanical modelling of bones, marine renewable energy and computation fluid dynamics of coal combustion and much more.

Neelofer BanglawalaApplications Developer, EPCCn.banglawala@epcc.ed.ac.uk

8

Michèle WeilandProject Manager, EPCCm.weiland@epcc.ed.ac.uk

NEXTGenIO www.nextgenio.eu

INTERTWinE www.intertwine-project.eu

ExaFLOW exaflow-project.eu

In 2008, DARPA (Defense Advanced Research Projects Agency, part of the US Department of Defense) published a report on the technology challenges of achieving exascale computing systems. In 2014, another report by the US Department of Energy described the top ten exascale system research challenges that still needed to be tackled, including energy efficiency, memory technology, programming systems and data management.

The first report predicted that exascale systems might be available from around 2015; as we were approaching that date, 2018 was beginning to emerge as the more likely timeframe. Today, it is widely accepted that 2020 (or more likely even 2022) are feasible timescales in which a usable exascale system will come online.

EPCC and exascaleWhat is EPCC’s role in this effort? We have always been at the cutting edge of HPC research, and when it comes to the exascale research effort, this is no different.

A couple of years ago, the European Commission issued a call for projects to address the core exascale research challenges. EPCC took this opportunity and is involved in three such projects (leading two of them), each with a different focus:

• NEXTGenIO, which is led by EPCC, looks at the impact of emerging non-volatile memory technologies on total system performance, with a particular focus on applications that rely on large amounts of data to be read or written either to storage or memory. NEXTGenIO’s use cases come from both a traditional HPC background (such as weather forecasting or computational chemistry), as well as targeting more novel HPC applications (such as machine learning or genomics).

• INTERTWinE is also led by EPCC, and focuses on improving the interoperability of popular parallel programming models. The aim is to allow programmers to choose (and mix) programming models that meet their needs, without sacrificing either productivity or performance.

• ExaFLOW investigates novel methods and numerical algorithms to optimise the performance and scalability of computational fluid dynamics applications. EPCC’s focus in this project is on efficiency, both in terms of performance and power/energy.

Now halfway through their lifetimes, the work on these projects is progressing well and EPCC remains an established leader in the world of HPC. But we won’t be resting on our laurels, so watch this space!

Outlook good for exascale

Centres of Excellence

Reaching the exascale has been a focus of the HPC community for several years, and EPCC has been a key player from the beginning.

In addition to the projects linked below, EPCC is also involved with two Centres of Excellence (CoE), which concentrate less on pure research and more on preparing end users and their applications for the exascale:

• BioExcel CoE supports both academia and industry in the use of HPC in biomolecular research (see opposite page).

• CompBioMed CoE works with the biomedical modelling community (see page 10).

Image: iStock.com/AlinaMD

9The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

I’ll start with the part of the project that I’m most closely involved with, which is reaching out to possible users of the Centre and ensuring its work will reflect the needs and interests of the computational biomolecular research community.

Interest groups The interest groups (IGs) – see my article in the previous issue of EPCC News – are growing both in their membership and activity:

• The Integrative Modelling IG now has its own page at bioexcel.eu where it highlights news and upcoming events specifically for this community.

• The Practical Applications for Industry IG has started regular phone meetings to discuss how best the Centre can support users from industry.

• The Hybrid Methods IG has been launched with a meeting in Sardinia, in conjunction with the European Joint Doctorate Programme in HPC for Life Sciences, Engineering and Physics, where a mix of early-career and established academics in the field met to discuss state-of-the-art techniques in this area.

In November, there will be an opportunity for all BioExcel interest group members to meet up in Amsterdam for the BioExcel

community forum. Registration opens soon, but you can sign up to our Interest Groups now to be kept informed.

Code developmentThe Sardinia meeting also offered the opportunity for BioExcel to introduce to the community new technical work that we have undertaken as part of the project to develop a new QM/MM interface for the code CPMD. This will make it easier to conduct simulations which take into account quantum effects to accurately represent how particles are interacting at atomic scale, but which also scale up to systems that are big enough to be of interest in real biological applications. As part of one of BioExcel’s use cases, this new interface is being evaluated using a real problem in collaboration with the Human Brain Project.

As well as CPMD, the project is also supporting the development of Gromacs and HADDOCK. These codes continue to be independently developed but BioExcel is helping to coordinate the code development, to share best practice between the code developers and to coordinate a number of BioExcel releases of the codes, the first of which has just been published.

Adam CarterProject Manager, EPCCa.carter@epcc.ed.ac.uk

Building a Centre of Excellence for Computational Biomolecular Research

The BioExcel project, which is establishing a Centre of Excellence in Computational Biomolecular Research, is now well into its second year. Here is an update on our progress.

Stem cells. Image: iStock.com/luismmolina

Find out moreOn our website you can watch webinars, join interest groups, sign up for training courses, download our software, ask questions, and have your say. www.bioexcel.eu

The BioExcel Centre of Excellence supports academia and industry with the use of advanced techniques for high-end computing.

10

CompBioMed: a new Centre of Excellence in Computational Biomedicine

Emily LumleyProject Manager, UCLe.lumley@ucl.ac.uk

Gavin J. PringleApplications Consultant, EPCCg.pringle@epcc.ed.ac.uk

EPCC is one of 15 core partners in CompBioMed. This user-driven Centre of Excellence in Computational Biomedicine will nurture and promote the uptake and exploitation of high performance computing within the biomedical modelling community. Its users come from academia, industry and clinical practice.

Computational methods based on human biology are reaching maturity in the biomedical domain, rendering predictive models of health and disease increasingly relevant to clinical practice by providing a personalised aspect to treatment. Computer based modelling and simulation is well established in the physical sciences and engineering, where the use of high performance computing (HPC) is now routine.

The CompBioMed Centre of Excellence prioritises innovation, promoting interdisciplinary entrepreneurial opportunities driven by our users’ needs. Its industrial partners have participated fully in the Centre’s activities from the outset, with the number of Associate Partners, currently 21, expected to grow continuously over the lifetime of the Centre.

CompBioMed will support and facilitate modelling and simulation activities, and provide education and training for a diverse set of user communities. The Centre will target research scientists from physical, computer and biomedical sciences, software and infrastructure

developers, from industry, and medical end-users, including clinicians. It will therefore invest in community building to spread knowledge, tools and best practice to students and researchers across this domain.

The Centre will provide a focal point for the development and sustainability of software tools and services, which can deliver high fidelity three- and four-dimensional (including time) modelling and simulation of all aspects of the human body, from the genomic level to the whole human and beyond, in health and disease.

HPC has the potential to enhance industries in the healthcare sector including pharmaceuticals and medical device manufacturers, underpinning a range of emerging sectors such as those concerned with e-health and personalised medicine. The innovative modelling and simulation techniques developed and promoted by our Centre will be of great interest and relevance to industrial researchers, HPC manufacturers and independent software vendors.

CompBioMed nurtures and promotes the uptake and exploitation of high performance computing within the biomedical modelling community.

11The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

EPCC and CompBioMedAs a Core Partner, EPCC is involved in many aspects of the CompBioMed Centre of Excellence. We enable CompBioMed members to access ARCHER, the UK’s National Tier-1 service. This includes access to the Research Data Facility (the UK-RDF service), which is ideal for the huge data files employed and the large pre- and post-processing demands of our users. Furthermore, CompBioMed users may employ Cirrus, EPCC’s new Tier-2 HPC service (see page 17), which is ideally suited for high throughput computing (HTC). HTC includes task farming jobs for ensemble simulations that are often found in biomedical simulations.

EPCC assists users to port their simulations to these platforms, and to enable the associated jobs to be managed remotely from a user’s own computer, ie HPC in the Cloud. This latter aspect is designed to enable medical practitioners to exploit the CompBioMed infrastructure without the need to be HPC-literate.

EPCC is also involved in this

Centre’s management and its associated training programme.

Further, we lead the Centre’s work package on “Innovation and Sustainability”. The purpose of this work package is to ensure the Centre of Excellence creates and implements both a strong sense of innovation and sustains its standing at the forefront of on-demand HPC for biomedical research.

Finally, Prof. Mark Parsons, Director of EPCC, is the Chair of the Centre’s Innovation Advisory Board, or IAB. The IAB will provide strategic guidance and support in Innovation Management within the project. The members of the IAB are drawn from stakeholder groups from within and external to the project to ensure a positive mix of perspectives and contributions: its composition will change to meet the specific needs of the emerging innovation activities as they evolve. The IAB will be called on for advice on both general and specific issues arising from Innovation Management activities, with specific focus on maximising impact within research, societal and commercial realms.

Modelling and simulation There are three key areas in which high fidelity modelling and simulation will be employed in the CompBioMed Centre of Excellence:

• Drug design

• Cardiovascular and musculoskeletal medicine

• Personalised medicine and clinical decision support.

The Centre has access to, and will develop software for, petascale supercomputers through our Core and Associate Partners. We subsequently plan to develop our work towards the evolution of hardware to the exascale.

The initial form of our Innovation Plan is now publicly available from the CompBioMed website:

www.compbiomed.eu/publications/deliverables

Research by the CompBioMed Centre of Excellence will lead to a computational human, a virtual workbench

for biomedical research. Graphic courtesy BSC.

12

John McCloskeyChair in Natural Hazards Science, School of Geosciences, University of Edinburghjmcclosk@staffmail.ed.ac.uk

Rob BaxterGroup Manager, EPCCr.baxter@epcc.ed.ac.uk

Building earthquake resilience in at-risk communities

Recent devastating earthquakes in Nepal and Italy have once again illustrated the need for better understanding, and more accurate operational forecasting, of aftershock sequences to assist emergency response: from risk assessment for civil protection and humanitarian NGO staff, to the planning of field hospital and temporary relocation sites, to logistical planning.

Seismology meets mobile phone

Funded by the UK’s Global Challenge Research Fund, REAR (Research for Emergency Aftershock Response) is a first step towards connecting earthquake science, the best contemporary methods in social science and the environmental humanities, with the revolution in digital communication and mobile phone technology to improve the resilience of communities in the developing world to major earthquakes. One of the key project partners is Ireland-based NGO Concern Worldwide, which plays a critical role by guiding the interaction between these developing ideas and the humanitarian NGO end-users.

It is well known in seismology that

aftershocks produce more fatalities and damage than would be expected from earthquakes of the same magnitude; they generate shaking in areas where much of the building stock has already been damaged, and where poor decision-making by the population can have fatal consequences. They also represent a significant impediment to emergency response efforts.

While we cannot reliably forecast individual main earthquakes, the statistics and physical science of aftershocks and operational forecasting has developed rapidly, and it is now possible to make actionable forecasts of the probability of both their location and the size, with full quantification of the uncertainties involved.

REAR aims to develop a simple mobile phone app that can be used by a network of volunteers in a given area, not to predict or measure main earthquakes but, given suitable placement (for example, in wall-mounted overnight charging stations) to enable mobile phones to be used as sensors which can measure the response of individual buildings to small seismic events. In turn, data recorded by the phone will be transmitted to a

Earthquakes have caused over three-quarters of a million deaths already in this century, and economic losses of over a quarter of a trillion US dollars since 1980, make them by far the most destructive of the natural hazards. EPCC has been involved in developing a new app that will lessen the danger of aftershocks.

Investment in protecting against earthquakes remains a low priority across much of the developing world. Until proven resilient construction becomes more available, high impact earthquake research must focus on evidence-based, targeted, socially tuned, resilience building; saving the maximum number of lives per dollar invested.

13The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

Ratna Khatri, 88 years of age, sits in the makeshift shelter her family have built since their home in the Dolakha District of Nepal

was destroyed by the earthquake, whilst her granddaughter goes through the box of supplies distributed by Concern Worldwide. Photographer: Brian Sokol/ Panos Pictures

for Concern Worldwide/Nepal/2015

central hub where an understanding of the risk to buildings can be assessed, and – hopefully – used to mitigate the effects of a future main quake.

EPCC’s role takes in both app development and the initial back-end data collection framework, but more than the technical angle the volunteer community approach is key to the success of REAR.

Correct community response to aftershocks reduces subsequent loss of life in earthquake-stricken areas but local response can be delayed or inhibited by social, cultural and political factors. This means that effective, large-scale user engagement with appropriate information, so essential in emergency response, requires investment in developing public awareness at scale, designing effective co-learning across multiple stakeholder groups, and building understanding of the issues that might limit or enable user engagement.

One of the hopes of REAR is that the network of citizen-science volunteers might use the custom designed, multi-function app to cooperate in the collection of data,

giving them ownership of the hazards assessment. The same app might create a community ready to receive resilience-building advice and education precisely tailored to their level of risk. The new data streams could then be used in current campaigns by the UN and NGOs to inform evidence-based reconstruction following a crisis.

REAR has assembled an inspiringly multidisciplinary team, many of whose members are working on developing-world issues for the first time. We are building foundations for the provision of near real-time operational forecasts of aftershock location and magnitude probabilities, based on dense networks of traditional seismometers. We are also exploring the global use of mobile phones, both as high density seismometer networks and as tools for community engagement and empowerment during emergency response.

The project will catalyse work with emergency providers and threatened communities to explore and exploit the social and cultural promoters and barriers to effective intervention. It’s truly exciting to be involved!

Read more on the EPCC blog:www.epcc.ed.ac.uk/blog

School of Geosciences: www.ed.ac.uk/geosciences

Concern Worldwide www.concern.net/

The CitiSeisApp interface showing information gathered from mobile

phone sensors.

14

Kostas KavousannakisGroup Manager, EPCCk.kavoussanakis@epcc.ed.ac.uk

NESS ends its journey into sound

With true inter-disciplinary focus, genuine user-engagement and over 75 publications overall, the project has been a great success for the University of Edinburgh, and for EPCC in particular.

In 2010, the School of Informatics funded a series of seed projects under the iDEA Lab initiative. One of these supported a collaboration between EPCC and Dr Stefan Bilbao at the School of Music (now in the Acoustics and Audio Group of the University of Edinburgh), to investigate the feasibility of employing graphics processor units (GPUs) to accelerate synthetic music codes.

Physical modelling synthesis

Digital sound synthesis goes back 60 years, when simple early algorithms were computationally efficient but unable to produce naturalistic sound.

Of particular interest to Stefan Bilbao is physical modelling synthesis, and especially finite difference time domain methods, where the acoustic space is represented as a grid, and the computation evolves with time. These methods can generate sound of great quality, they allow the development of models that may represent real or imaginary instruments, and can be used to also model accurately sound travel in 3-D spaces.

The main drawback of these methods is their computational complexity. This affects the applicability of some instrument-models, where real-time execution is very desirable, and it also makes 3-D modelling expensive.

NESS is born

With the iDEA Lab pilot showing promise that the codes could be ported to parallel architectures, Stefan Bilbao was successful in securing NESS a prestigious five-year European Research Council fellowship.

The project was set up with three teams. The algorithm team at AAG developed models for instrument families like brass, percussion and strings, and for 3-D spaces. Their emphasis was on creating portfolios of models that can be used for sound synthesis and room modelling, with some work also on algorithms that are more computationally effective. The acceleration team at EPCC ported the Matlab models into C++, optimised and accelerated them using various techniques and types of hardware, and made them available through a simple user interface. Finally, NESS had funding for visiting composers, who made use of these accelerated models - and of the studio that Edinburgh College of Art made available to the project - to develop new pieces of music.

Code acceleration

NESS was very interesting for EPCC. One of the findings of the project was that some of the instrument families, eg brass, do not need much in the way of acceleration, and can run faster than real-time even on our now

The Next Generation Sound Synthesis project (NESS) has concluded its five-year journey to investigate the feasibility of employing graphics processor units (GPUs) to accelerate synthetic music codes.

15The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

The NESS project is supported by the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement number StG-2011-279068-NESS.

Part of the illustration “Example of directionally-dependent dispersion error on cubic, FCC, BCC

grids”, which appears on the Ness website.

For more information – and sound samples – see the NESS website:

www.ness-music.eu

My personal highlight was realising the value that NESS has added to the day-to-day work of our composers. I will always remember listening to the first NESS-based piece, Gordon Delap’s “Ashes to Ashes”, when it was performed publicly for the first time in the NESS studio.

superseded, older server. Others, eg the snare-drum code, were notoriously non-conducive to sufficient acceleration, and research was carried out by the modelling team to replace some of the algorithms used.

Instruments

The true showcase of EPCC’s contribution to the acceleration effort was the multi-plate code. As published in 20151, James Perry of EPCC employed a combination of GPU, multicore, vectorisation, banded matrices, merging of operations, and mixed precision to accelerate the code. Depending on the size of the problem, it is still around 240 times slower than real time, but it is between 60 and 80 times faster than the original, unoptimised Matlab code.

3-D rooms

EPCC contributions to the 3-D rooms code were of a slightly different nature.

EPCC’s Paul Graham ported the code to MPI and ran various demonstrative examples of how sound behaves in big rooms, using ARCHER, the UK national supercomputer. The code runs slower than on the project’s GPU server, but much bigger rooms are possible because of the sheer amount of available memory on ARCHER. The amount of computation required is significant: a 16x16x16 m3 room model, sampled at a rate of 44.1kHz, requires 28 GB of memory and it

takes 240 processor-cores of the order of 80 minutes of wall-clock time to produce one second of audio output.

The NESS website includes some of the sounds generated, but these stereo renditions do not do justice to the excitement the mosquito model generates on an eight or 16-channel set-up.

Project legacy

NESS was very successful and allowed us to enhance the GPU and numerical modelling skillset at EPCC. The collaboration between the teams was excellent, with the models iterating between the algorithms and acceleration teams before being made available to the composers. EPCC’s Alan Gray was the second supervisor for two of the associated six NESS PhDs and we are happy that, of the 77 NESS publications to date, seven involved EPCC.

It is easy to get carried away with the endless possibilities for sound synthesis, but where we see real room for growth is the 3-D rooms. These models will need to be improved, refined and validated, but they have the potential for a tremendous impact in various areas of civil engineering.

This is where the University of Edinburgh excels, building highly skilled, cross-disciplinary research teams that can influence people’s everyday life. I look forward to future opportunities to build on the success of NESS.

1. J. Perry, A. Gray and S. Bilbao. Optimising a Physical Modelling Synthesis Code Using Hybrid Techniques. Proceedings of the International Conference on Parallel Computing, Edinburgh, September 1-4, 2015.

16

These processor varieties represent the two ends of the processor spectrum we see in modern computing platforms. Xeon is the multi-core processor line used in desktop and server class systems; Xeon Phi is the many-core processor line for computational simulation applications.

The primary difference between the two types are the number of compute cores on the chip (60-70 for Xeon Phi, 10-20 for Xeon), and the support for floating point calculations. The Xeon Phi processor supports very large vector instructions (performs many floating point calculations in a single instruction cycle), with up to two 512-bit wide vector units (8 double precision numbers) per core; the Xeon processor supports single 256-bit or 512-bit vector units.

If we optimise the use by an application of these large vector units on a Xeon Phi processor, we should also get benefits on the Xeon processors. It is possible to vectorise applications by hand (eg using intrinsic instructions), but our focus has been on restructuring codes to ensure the compiler can vectorise the code effectively.

Improving vectorisation

One example of such work was a collaboration with Dr Angus Creech of the Institute for Energy Systems in Edinburgh, where we looked at improving the vectorisation of the CFD modelling package Fluidity.

Many of the key computational kernels in Fluidity are generalised to allow different types of simulation. However, this impacts performance as it inhibits the compilers’ ability to optimise these routines.

To optimise the routines for tidal simulations, more specialised versions were created that converted dynamic array allocations to static allocations, added compile time loop length definitions to enable the compiler to vectorise the loops, and inline the main computational routines.

This was combined with code that selects optimised routines at run time if the correct type of simulation is being run, and uses the original code for functionality that hasn’t been optimised. This enabled an approximate two-times speedup of the whole code, with simulations completed twice as fast as before.

Future plans

In this fourth year, we will be looking at similar applications and optimisation work, including optimising the statistical programming language, R, for the Knights Landing processor; investigating performance portability for applications across Xeon, Xeon Phi, and GPU processors; and developing models to understand the best ways for applications to use the Xeon Phi’s high bandwidth memory.

If you’re interested in the Xeon Phi processor, in how your application may perform on the hardware, or any other aspect of this work, please don’t hesitate to get in touch:

a.jackson@epcc.ed.ac.uk Twitter: @adrianjhpc

Intel Parallel Computing Centres: https://software.intel.com/en-us/ipcc

Fluidity project: http://fluidityproject.github.io

Adrian JacksonResearch Architect, EPCCa.jackson@epcc.ed.ac.uk

The Intel Parallel Computing Centre at EPCCWe are entering the fourth year of the Intel Parallel Computing Centre (IPCC). This collaboration on code porting and optimisation has focussed on improving the performance of scientific applications on Intel hardware, specifically its Xeon and Xeon Phi processors.

An image produced by the CFD modelling package Fluidity.

17The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

EPCC has used this grant to extend Cirrus, its SGI ICE XA system, transforming it into a major multi-core HPC system with 10,080 state-of-the-art Intel Broadwell cores, and increasing its storage to 2.4PB. 2PB of this storage is a new Tier 2 Research Data Facility built around a DDN WOS object store.

Increasing HPC capability and uptakeThe new Tier-2 centres provide a key part of the UK HPC ecosystem, bridging the gap from Tier-3 (institutional HPC systems) up to Tier-1 (national HPC systems, such as ARCHER which is also hosted and managed by EPCC).

While many universities might be expected to host a Tier-3 system, Tier-2 is beyond the capabilities of most. This investment by EPSRC will increase the range of computing capability and capacity on offer to

UK researchers. The range of technologies on offer through the EPSRC Tier-2 facilities complement the UK National Supercomputing Service, ARCHER, and allow a wider range of research communities across the UK to use HPC to further their research.

This tiered approach to HPC is part of EPSRC’s core strategy for research computing.

Composite Tier-2 systemIn addition, EPCC is taking a leadership role by providing free use of its SAFE system administration software, which supports easy user management and reporting on advanced computing services. From a user management point of view, this will allow the new Tier-2 centres across the UK to operate as a single, joined-up system.

“EPCC is very proud to have won this funding which has allowed us to hugely expand our Cirrus system. This will greatly benefit our scientific researchers and provide our industry users with a much larger resource.” Mark Parsons,EPCC Director

EPCC has received £2.4m from the Engineering and Physical Sciences Research Council (EPSRC) as part its £20m investment in six new Tier-2 HPC centres.

Cirrus transformed into Tier-2 system

EPCC’s Cirrus system. Callum Bennetts, Maverick Photo Agency

Andy TurnerProject Manager, EPCCaturner@epcc.ed.ac.uk

18

Masters programmes at EPCC

As part of our MSc Programmes EPCC organises a yearly series of guest lectures given by external speakers from industry and academia.

These are formally part of the “HPC Ecosystem” course and like that course as a whole they complement the core technical skills taught on our other MSc courses by offering insights into how high-performance computing actually operates in the real world.

The lectures strengthen our students’ awareness and

understanding of the evolving HPC landscape and its key players including HPC hardware and software vendors, service providers, user communities, etc and how these relate to each other. Students also find the lectures offer inspiration for possible career directions and opportunities in HPC.

The MSc guest lectures are aimed at our MSc students but are also attended by our PhD students, EPCC staff, and are usually open to interested externals.

Annual Guest Lectures series

We offer two well-established MSc degree programmes: the MSc in High Performance Computing, and the MSc in High Performance Computing with Data Science.

Enrolling in an MSc degree programme with EPCC translates to making a significant investment in one’s professional career.

With the option to study flexibly, our programmes enable students to gain a specialist qualification in areas of computing that are in great demand worldwide. Both programmes have a strong practical focus and provide access to our leading-edge HPC systems.

Edinburgh is home to the brightest tech start-up scene outside

London, and this means that our students benefit from unrivalled industry placement opportunities. See opposite for two of this year’s examples.

We are excited that our programmes will have their new home in the Bayes Centre. We anticipate that the new facilities will not only allow students a more beneficial academic experience, but provide greater co-curricular opportunities that are so important to the student experience.

Both our programmes offer flexibility with a rich choice of units and excellent job prospects on graduation.

Image: iStock.com/massimofusaro

Guest Lectures archive Abstracts and slides from Guest Lectures from 2009–2016 can be found on our website:www.epcc.ed.ac.uk/msc/programmes/why-study-with-us/meet-our-collaborators

www.epcc.ed.ac.uk/msc

Arno ProemeApplications Consultant, EPCCa.proeme@epcc.ed.ac.uk

Ben MorseMSc Programmes Officer, EPCC b.morse@epcc.ed.ac.uk

19The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

Optos Plc is a local Nikon company that produces scanning devices that create high-resolution images of the retina, which are used to diagnose eye diseases.

Optos’ devices use FPGAs as an interface between the device and a CPU, and the FPGA performs image pre-processing while the data are accessible to it. To take advantage of the benefits that GPUs bring to image processing, such as core count and high bandwidth memory, the pre-processed data are transferred to a GPU.

Our MSc student is working on a project that aims to produce means of transferring the data directly from the FPGA to the GPU, without going via the CPU. Although the project is just starting, an initial exploration suggests the approach will be to

create drivers for an FGPA to write the data straight to the GPU using remote memory access.

The overall intention is to reduce the time between the initial scan and a health-care professional viewing the image: the longer-term aim is to reduce this time sufficiently that the practitioner can scroll through several images ‘while the patient is still in the chair’.

The project is being undertaken by an international student with an undergraduate degree in Information Systems Engineering. During the taught part of the MSc he gained the programming skills required to complete this project. Working on this project means he will gain insight into ground-breaking medical technologies, thereby increasing his employability.

Optos Plc

Mallzee is a shopping platform that allows users to indicate items they like or dislike. The latest version of the app also allows users to buy an item with a few more clicks.

While users can choose from a wide range of products, retailers and designers have to deal with increasingly selective customers. Mallzee wants to show its users the products they are more likely to prefer, while also giving retailers useful information about customers’ preferences so they can design suitable products.

With the help of two MSc students from EPCC, Mallzee is tackling both problems.

Every swipe collected from hundreds of thousands of customers, plus the attributes of the item that was liked/disliked and the attributes of each customer is added to the database that will be used for both projects.

In the first project, Machine Learning techniques will help predict the number of sales at an individual product level, helping the retailers in their business decisions. In the second project, a recommender is being built to enable Mallzee to show customers products they are more likely to find appealing and to buy.

MSc student placements with local industryOur industry-based dissertation projects enhance our students’ skills and employability by tackling a real-world project, gaining workplace experience, exploring potential career paths and building relationships with local companies.

Marc SabateApplications Consultant, EPCCm.sabate@epcc.ed.ac.uk

Mallzee

Mark TuckerApplications Consultant, EPCCmark.tucker@epcc.ed.ac.uk

A child’s healthy retina. Image: Optos.

Find out more

www.epcc.ed.ac.uk/msc/programmes/programme-structure/dissertation

Adarsh Janakiraman – a graduate of the MSc in HPC with Data Science – receiving the 2015/16 Class Medal for his industry-based dissertation.

Adarsh was based in Mallzee, a company we are working with again for this year’s MSc.

20

Going the distance: EPCC presents its first supercomputing MOOC

Weronika FilingerApplications Developer, EPCCw.filinger@epcc.ed.ac.uk

Course structure and delivery In accordance with the MOOC methodology of presenting the content in small, easily digestible portions, we designed this course to last for 5 weeks.

Each week has a distinct theme, and is further divided into smaller modules called ‘activities’, consisting of a number of ‘steps’. Steps are the smallest units of the course structure and, regardless of their type – article, video, discussion, exercise, quiz or test – should not require more than 20 minutes to complete. Learners can spend as much or as little time on each step as they wish, and do them at any time. In our estimate the week’s worth of content should not take more than 3 hours to complete. Learners are granted access to all of the material at once, which allows them to proceed at their own pace.

Programme contentThe first week provides a gentle introduction to the world of supercomputing, including some

basic terminology, a brief historical overview and descriptions of some of the state-of-the-art machines.

The second week focuses on the physical components of supercomputers and how they compare to modern laptops or game consoles. At the end of this week, students should appreciate why we use thousands of standard CPUs rather than creating one super-CPU that is thousands of times faster.

In week three we discuss how to make use of all of these computing cores, covering topics of parallel programming and parallel performance.

Week four is about computer simulations, with weather modelling used to illustrate the concepts and life cycle of computer simulations.

Finally, week five presents three examples of research conducted on supercomputers: Illustris, the biggest cosmic simulation ever; the Blue Brain project, an attempt to create a digital reconstruction of the brain; and discovering new materials. We also talk briefly about possible future developments.

Massive Open Online Courses (MOOCs) provide free web-based distance learning opportunities to large numbers of geographically dispersed students. Here at EPCC we are always keen to talk about supercomputing, and becoming involved in this MOOC was a natural development for us.

“Thank you for providing this very interesting and engaging course. The content of education material was brilliant.”

“Thank you for an excellent course. The information provided was very educational and I have learned so much from the modules and other participants.”

www.futurelearn.com/courses/supercomputing/

21The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

Course participants This course was designed to be accessible to as many people as possible, and participants do not require any specific computer knowledge. However this does not preclude more computer-literate learners from participating and enjoying it.

Looking at the self-introductions and various comments, we can distinguish three distinct groups of learners: current/former IT professionals, post-graduate students and academic researchers making use of HPC resourses, and the general public.

This diversity led to some very interesting conversations between significant numbers of learners. The discussion topics ranged from very technical and in-depth, through general interest questions, to philosophical and ethical considerations. Although not all learners were active in the comment sections, many noted that they benefited from following other learners’ conversations.

3,263 people registered for the

course and 1,721 people accessed the content, which is a typical conversion rate for FutureLearn. Due to its rather specialist nature, ours was quite a small course by FutureLearn standards. However, those smaller numbers allowed for more interactions between the educators and learners and, possibly more importantly, between learners.

How did we do?Feedback from people who completed the course was very positive and we look forward to running it again. Comments included:

“Thank you for making this course available, it has been a huge insight into my future prospects and has taught me not only about huge supercomputers, but also about the fundamentals of computer science itself!”

“Thank you for a thoroughly enjoyable course. This was a topic about which I knew nothing before starting the course and so I have learnt much. The future looks interesting.”

Two examples of scientific research made possible thanks to modern supercomputers.

Top: The biggest cosmic simulation ever. The Illustris project was run on 8,192 compute cores, and took 19 million CPU hours – the equivalent of one computer CPU running for 19 million hours, or about 2,000 years!

http://www.illustris-project.org

Bottom: The Blue Brain Project is building biologically detailed digital reconstructions and simulations of the rodent, and ultimately the human, brain. The supercomputer-based reconstructions and simulations built by the project will offer a radically new approach for understanding the multilevel structure and function of the brain.

http://bluebrain.epfl.ch

Registration

The date of the next run is planned to be at the end of August 2017. To be notified of forthcoming runs, please register on the course page below.

We hope you will join our next run!

Registration: www.futurelearn.com/courses/supercomputing

This MOOC was developed by EPCC in collaboration with SURFsara (Netherlands) on behalf of the Partnership for Advanced Computing in Europe (PRACE). The course is hosted on FutureLearn, The Open University’s digital education platform.

22

Toni CollisApplications Consultant, EPCC a.collis@epcc.ed.ac.uk

Women in High Performance Computing (WHPC) will make its third appearance at ISC with activities designed to spread the message of the value of a diverse HPC community, and to open discussion on the topic. We also want to address some of the barriers we all face in building a more inclusive, diverse workforce.

To bring about change, it is important to engage as many people as possible: women need to be provided with the network and opportunities that they may be missing, and everyone needs to understand the barriers to participation for women and how to address them. This is not a simple problem: if it were, we would have equal representation of the genders across all areas of science, technology, engineering and mathematics employment as there are plenty of leaders, managers and initiatives trying to address it.

At ISC 2017 we will encourage discussion of the difficulties faced both by female employees and their employers. By hosting this discussion we hope to build interest in the vision of WHPC, ensure everyone involved understands the benefits of promoting and achieving inclusivity and, specifically, the importance of increasing the participation of women.

Birds-of-a-Feather

Our Birds-of-a-Feather (BoF) on ‘Practical steps to diversifying the HPC workforce’ will discuss

unconscious bias, impostor syndrome and stereotype threat, and finish with a session for generating ideas, encouraging attendees to tackle difficult situations and understand the underlying causes for the under-representation of certain groups.

Panel discussion

We will host a panel discussion on diversifying the HPC workforce. One of the key aims of the WHPC workshop series has been to give women the opportunity to present their work, develop networks and gain valuable workplace skills. This will continue at ISC with a series of posters and lightning talks by women, followed by short talks by leaders in the HPC workforce on career skills and development.

Networking events

Our ISC activities will also support personal networks of peers, potential mentors, and future employers. We will host a speed networking session during the workshop and a networking luncheon to showcase the work we are doing and, crucially, bring in business and academic leaders to meet women working in HPC.

Our overall aim in participating in the ISC High Performance 2017 conference is to provide a range of activities designed to reach as broad an audience as possible, to prompt discussion, and encourage the adoption of best diversity practices throughout the HPC community.

Women in HPC will offer a series of events at ISC this year designed to broaden women’s participation in the conference and improve diversity in HPC.

Women in HPC at ISC 2017

Most of us recognise that diverse teams are good for productivity and output. But do you know how to improve diversity and build a more inclusive environment? Have you heard of unconscious bias, stereotype threat or imposter syndrome? WHPC works to understand and combat such barriers to a diverse HPC workforce.

More on WHPC’s activities during ISC can be found at: www.womeninhpc.org/women-in-hpc-at-isc-2017

See also the article opposite.

We have put together a range of activities to encourage and enthuse everyone in the HPC community to increase the participation of women.

The Women in HPC networking lunch at ISC 2016.

23The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

NEXTGenIO Tutorial 01: Understanding and Improving I/O performance on HPC systems. Co-presented by Adrian Jackson. (June 18)

2017 Student Cluster Competition Twelve teams, including one comprising students of our MScs, will take part in the three-day Student Cluster Competition. EPCC team advisor: Manos Farsarakis. (June 19-21)

Vendor Showdown 02 Joint Moderator: Mark Parsons (June 19)

INTERTWinE Poster: MPI RMA as a directory/cache interoperability layer. By Nick Brown. (June 20)

NEXTGenIO BoF: Towards addressing Exascale I/O requirements and challenges. Joint speaker: Adrian Jackson Chair: Mark Parsons (June 20)

Energy Efficiency in HPC Panel: Moderated by Michèle Weiland (June 21)

Women in HPC BoF: Practical Steps to Diversifying the HPC Workforce Joint speaker: Toni Collis (June 21)

PRACE Training Workshop Presentation: SME HPC Adoption Programme in Europe. By Paul Graham (June 22)

Women in HPC 6th International Women in HPC Workshop. (22 June)

WOPSSS workshop Invited talk: Non-volatile memory for next generation I/O. By Michèle Weiland (22 June)

Our leadership of the Fortissimo project, helping the European manufacturing industry access on-demand HPC resources.

Our involvement in the Centres of Excellence initiative via BioExcel and CompBioMed, enabling use of HPC and HTC in life sciences.

Our push towards the exascale goal through FETHPC projects, including leadership of NEXTGenIO, ExaFLOW, and INTERTWinE.

Our commitment to increasing diversity through founding the Women in HPC network.

Our mini supercomputer, Wee Archie, will be on our booth, and we’ll be talking about how we use it in our public outreach programme. You can also ask us about the training and education opportunities at EPCC, including our two MScs, online learning options, PhDs and training for industry.

Our scheduled activities

L to R, from top left: Nick Brown, Toni Collis, Manos Farsarakis, Paul Graham, Adrian Jackson,

Mark Parsons, Colin Paul, Michèle Weiland, Mirren White.

EPCC booth: B1223

We will take part in a number of events at ISC17, and will also host a booth on the exhibition floor. Come and find us to hear about what we’ve been up to – and what’s to come!

EPCC at ISC 2017Frankfurt, June 18-22, 2017

Meet our team at ISC!Come to our booth to learn more about:

Mirren WhiteProject Dissemination and Exploitation Officer, EPCCm.white@epcc.ed.ac.uk

24

Gordon GibbApplications Developer, EPCCg.gibb@epcc.ed.ac.uk

Our Wee Archies have been to various events throughout Europe and the United States in the past year, including large science festivals, conferences, and even an art installation.

Wee Archie was created to give children and adults a hands-on experience of a ‘real’ supercomputer. Built for portability and ease/speed of set-up, all its components are housed in a single clear acrylic casing, with only a power cable and an Ethernet port as inputs. The clear case gives an unobstructed view of the internal components, allowing the inner workings to be easily shown.

Wee Archie was designed to be a supercomputer analogue, and contains service nodes, compute nodes, a power supply, a shared filesystem and an interconnect, so giving an insight into the structuring of real-life supercomputers.

Demonstration codes

We have created several demonstration codes to engage users of Wee Archie, with the

particular aim of showing some of the science being researched on ARCHER, the UK’s supercomputing service. Previously we have used a demonstration based on Gaitsym, where users could design their own dinosaurs by adjusting the body and limb sizes of various templates, then race them. We also have a demonstration using the NAMD molecular dynamics code, where a pheromone must be guided out of a protein in mouse urine!

CFD demo

Another important use of ARCHER is computational fluid dynamics (CFD) research. To represent this on Wee Archie, we created a demo which allows a user to design the cross section of an aeroplane wing, whose aerodynamic properties are then determined by Wee Archie. Afterwards, the user can find out how far their aeroplane could fly on a single tank of fuel.

With this new CFD demo, users are first shown the cross section of their aeroplane wing, and they can adjust the angle of attack, camber (curve),

Wee Archie is our miniature model supercomputer made from Raspberry Pis. It has been such an invaluable resource for our public outreach activities that we had to build a second one!

Wee Archie gives you wings!

Our Wee Archie mini-supercomputer and a boy called Archie. Pic: Callum Bennetts,

Maverick Photo Agency.

25The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

and thickness of the wing using sliders to create their desired shape. They then submit this shape to Wee Archie, which will calculate the airflow over the wing. While this calculation is being performed, the client displays dashed lines over their wing to show how the computational domain is split up between the different nodes on Wee Archie. The calculation takes around 25 seconds.

Once the simulation is complete, the airflow around the wing is displayed on screen, along with a vector indicating the direction of the net force (lift and drag) on the wing. The air pressure around the wing can also be displayed, presenting the possibility of a conversation about what conditions are needed in order to generate lift.

Now for the fun! The user can click on the “take off” button, which will attempt to make a plane fly with that wing. An animation shows the plane either soaring gracefully into the air, or falling into the sea.

In addition to this, the user can see how far their plane can fly. This is

displayed as an animation, with a shaded area on a world map growing larger as the fuel gauge empties. Users can compete with each other to see whose wing can carry the plane the furthest. If a plane cannot take off, it could still have a large range because in reality a plane’s wing shape is not fixed, but alters during take-off, cruise and landing to optimise its shape for the task at hand. Users can experiment with different wing designs, and determine the required properties for takeoff and cruising. This educates the user on aerodynamics, and also demonstrates why a supercomputer is useful for designing and testing many prototypes in quick succession.

We plan to create several other demos in future, showcasing more kinds of research carried out on ARCHER. We (and Wee Archie) hope to see you soon at a future science festival or conferences!The Demo Framework was written by Gordon Gibb and Alistair Grant. The CFD Demo was written by Gordon Gibb, the artwork was by Mirren White, and Kara Moraw added some enhancements to the demo.

Wee Archie is constructed out of 18 Raspberry Pi 2 model B boards (Raspberry Pi 3 model B in the case of Wee Archie 2), connected via Ethernet, with a shared filesystem via NFS. The clusters were designed and built by Alistair Grant (EPCC), Peter Reid & Mark Reynolds (FUSION).

To facilitate demonstrations, we developed a Python framework, which allows a server to run parallel simulations on Wee Archie, and graphical clients for designing inputs and viewing outputs on laptops. The framework provides a template from which new demonstrations can be easily created.

More information on Wee Archie can be found at: www.epcc.ed.ac.uk/discover-and-learn

The Wee Archie CFD demo. Left to right, from top: 1. Users can change the shape of their wing. 2. Wee Archie has calculated the airflow around the wing, as well as its aerodynamic properties. 3. The dashboard. 4. The plane takes off. 5. The plane crashed into the sea and passengers must board the life rafts! 6. The range of

the plane is displayed.

.

26

Mario AntoniolettiSoftware Architect, EPCCm.antonioletti@epcc.ed.ac.uk

Wee Archlet: our homemade Raspberry Pi cluster

When we started our outreach activities at EPCC, we wanted to allow the general public to use an application running on a real supercomputer. We developed an appropriate molecular dynamics-based application that actually ran on ARCHER, the UK national supercomputing service, which is hosted by EPCC.

However, although it ran a lot faster than on a laptop, the activity didn’t convey the experience of working with a supercomputer or give the user new information. In short, no wow factor.

The solution was to create a representative self-contained supercomputer that we could take to events and which would show the principle behind the parallelism that underpins supercomputing. We were able to do this when we obtained funding through the Engineering and Physical Sciences Research Council, and so Wee ARCHIE was born.

Based on 18 networked Raspberry Pi’s, Wee Archie has an attractive set of LED lights that convey some of the excitement of supercomputers. After all, Thinking

Machines had done it with their CM-5 that appeared in the film “Jurassic Park”, so why not us? Moreover, we could run our previous demonstrators on this system, possibly not as fast as on the actual supercomputers, but enough to demonstrate the underlying principles that larger systems employ.

But you don’t need the fancy packaging of Wee ARCHIE to develop a viable Raspberry Pi cluster. Normal Raspberry Pis with simple packaging such as brick casing (for example LEGO), can be used instead - we call ours Wee Archlet.

Learning how to connect all of the elements together to form a clustered network is interesting, and the resulting system can be used to explore how supercomputers work.

To support this, we have produced a set of instructions for how to build a Raspberry Pi-based cluster. See: http://bit.ly/2qLEILx

You can contribute to these instructions yourself via GitHub or note any issues here: http://bit.ly/2rsQATm

Download our instructions to build your own Raspberry Pi cluster.

I would love to hear the experiences of anybody who builds their own Raspberry Pi cluster using our instructions!

Contact me at: m.antonioletti@epcc.ed.ac.uk

The pictures show examples of our Raspberry Pi clusters.

27The newsletter of EPCC, the supercomputing centre at the University of Edinburgh

Our booth offered three activities:

Beanbag sort, an activity that aims to show that the speed at which beanbags can be colour-sorted increases as more people become involved. We use it to illustrate how parallel computing employs multiple cores to solve a problem.

Supercomputing app, a game developed at EPCC that allows participants to build and manage their own computing centre.

Wee Archie and Wee Archlet. Wee Archie is the smaller sibling of

ARCHER: a cluster of 18 networked Raspberry Pis. Wee Archlet is even smaller, with five Raspberry Pis (see page opposite). Both systems are used for educational purposes. A aeroplane CFD demo was specifically developed for use at this event (see page 24).

Four to five EPCC staff were available on each of the four days to introduce HPC to a predominantly young audience. All of us who participated in Fair enjoyed it, although it was exhausting, and we look forward to attending next year.

EPCC at the Big Bang Fair

The three-day AAAS conference brings together scientists, engineers and the general public from across the world to discuss latest developments. The UK presence included Wee Archie, and we premiered our aeroplane CFD demo (see page 24) at the event.

Building on earlier work that used Wee Archie to model airflow over a wing and display the results, this demo has been extended to compute many of the parameters crucial for flight. Participants can then view their wing on a passenger aircraft, see how easily the aircraft will take off (or not!), and the overall range of the plane on a single tank of fuel. By then modifying the wing parameters and re-simulating on

Wee Archie, the wing can be optimised for specific aspects, but with a trade-off between take-off performance and range.

I also gave a short general interest talk introducing supercomputing to the general public. I spoke about how computational science has a massive impact on our lives in ways which many don’t realise, and explained that some really exciting work in this area is being done in the UK.

Wee Archie was very popular and constantly busy throughout the show, sparking many interesting discussions about HPC and its role in UK science and engineering.

Nick BrownApplications Consultant, EPCCn.brown@epcc.ed.ac.uk

Wee Archie goes to BostonNick Brown and Wee Archie, our Raspberry PI mini super-computer, went to Boston in February for the annual American Association for the Advancement of Science (AAAS) meeting.

Juan Rodriguez HerreraApplications Developer, EPCCj.herrera@epcc.ed.ac.uk

EPCC made its second appearance at the annual Big Bang Fair in Birmingham, which offers exhibitions and workshops on science, technology, engineering and maths for schoolchildren from across the country.

The EPCC stand at this year’s Big Bang Fair.

Learn with EPCC at the University of Edinburgh

Master’s degrees in High Performance Computing (HPC)

and in HPC with Data Science

EPCC is the UK’s leading supercomputing centre. We are a major provider of HPC training in Europe, and have an international

reputation for excellence in HPC education and research.

Our two MSc programmes have a strong practical focus and provide access to leading-edge HPC systems such as ARCHER, the flagship UK National Supercomputing Service, and Cirrus,

one of the new EPSRC Tier-2 National HPC Facilities.

Through EPCC’s strong links with industry, all students are offered the opportunity to undertake an industry-based dissertation project.

Both programmes will be located in central Edinburgh at the Bayes Centre from September 2018.

The University of Edinburgh is consistently ranked among the top 50 universities in the world.* * Times Higher World University Ranking

www.epcc.ed.ac.uk/msc

Study HPC in the heart of Edinburgh