UK News from CERN Issue 82

 

Issue 82 contents

SKA-mid Africa

Array of ideas for science data

Read article

CERN has signed an agreement with the SKA Organisation to formalise collaboration in extreme-scale computing.

Adam Bozson

Unfolding ATLAS data with A Bozson

Read article

We’re living in an age of intelligent machines; algorithms and machine learning are part of our world.

Hannah Short

Computer security is all about communication

Read article

The days of multiple passwords might be on the way out, but behind the scenes it isn’t so simple.

Cross bay infill

Data racks up at CERN

Read article

UK company Dataracks has supplied equipment to CERN’s new network hub.

CERN Data Centre

Inside story

Read article

Want to go inside the CERN Data Centre? On a racing drone? Of course you do!


 

Array of ideas for science data

SKA-mid Africa

SKA-mid Africa – close up artist impression
(Credit: SKA Organisation)

 

CERN has signed an agreement with the SKA Organisation to formalise collaboration in extreme-scale computing.

The Large Hadron Collider and the Square Kilometre Array will be the two largest producers of science data on the planet, and a new agreement puts in place a formal framework for collaborative projects that will look at issues that the two projects have in common; exascale computing and data storage.

The SKA will be the world’s largest radio telescope with an array of antennae spread across southern Africa and Australia. Phase one of the project (representing approximately 10% of the whole of the SKA) is due to start scanning the Universe in 2020 and will generate around 300 petabytes (PB) of data every year. This is ten times more than today’s biggest experiments.

The LHC experiments have only just reached 200 PB after seven years of operating. But when the LHC is upgraded to deliver higher luminosity beams in 2020, the experiments are expected to generate more than 200 PB each year.

Common areas of collaboration include data acquisition, storage, management, distribution and analysis.

One particular area of interest is the development of a Science Cloud. “CERN has proposed the concept of the Federated Open Science Cloud with other EIROForum members,” says Ian Bird (CERN) Project Leader for World-wide LHC Computing Grid. “This agreement is an important step in this direction. Essentially, we will provide a giant cloud-based, Dropbox-like, facility to science users around the world, where they will be able to not only access incredibly large files, but will also be able to do extremely intensive processing on those files to extract the science.”

As part of the agreement, CERN and SKA will hold regular meetings to monitor progress and discuss strategy. There will also be joint workshops, projects and prototypes to investigate concepts for managing and analysing exascale data sets in a globally distributed environment.

 

Contents


 

Unfolding ATLAS data with A Bozson

Adam Bozson

Adam Bozson
(Credit: A Bozson)

 

We’re living in an age of intelligent machines; whether it’s facial recognition software on Facebook, voice activated services on your smartphone or the technology for driverless cars, complex algorithms and machine learning are part of our world.

PhD student Adam Bozson (RHUL and ATLAS) is using similar technology to improve the accuracy of measurements by the ATLAS experiment.

“Collisions happen in the heart of the machine with the particle debris being recorded in the layers of detectors that surround the interaction point. The direction and energy of each particle has to be reconstructed to understand the collision,” he explains.

“What makes the reconstruction process much more complicated is the detector itself; all the materials used in its construction, as well as its mass and the energy used to operate it, deflect the collision debris in different ways.”

Adam is aiming to deliver a set of collision data with all the effects of the detector removed. In physics, the process is called ‘unfolding’ and it matters because it enables theorists to compare and combine results from different experiments to confirm existing theories or look for new physics.

The first and most resource intensive step in the process of unfolding is to develop the best possible simulation of the detector. The deflection effects of the different parts of the detector are random, but it’s possible to spot trends. “The simulation is never going to be perfect, but we want to get it as close as possible,” explains Adam. Perfecting the simulation is a never-ending process – any changes to the detector or the way that it operates need to be incorporated into the simulation immediately.

Combining theoretical models of collisions with the detector simulation gives physicists a set of data that, subject to all the uncertainties of the simulation and the uncertainties of nature itself, they should expect to see from the real collisions.

These results are used to train a machine learning algorithm to ‘subtract’ the detector simulation from the actual data observed by ATLAS. The unfolded results are then compared with the theoretical models to see how closely they match.

What makes Adam’s machine learning approach better than any previous techniques is that it can handle the uncertainties in a proper mathematical manner, minimising further uncertainties.

“My PhD brings together physics, statistics and computer science,” says Adam. “Having a solid understanding of the experimental hardware is essential, but so is a passion for solving an interesting problem; I want to find out what’s going on at the heart of nature.”

Having the skills to develop the computer code is important, but the right hardware matters too. Thanks to the popularity of computer gaming, high quality graphics technology is now cheaper and more accessible. “For calibrating jets [a spray of particle debris from a collision] the algorithm is 50 times faster because we’re using graphics cards,” says Adam. “GPUs suit this type of analysis because they can process lots of small calculations in parallel really quickly.”

Using machine learning to reconstruct particle collisions is a very specific application, but the core technique is very similar to that used by industry. It’s difficult to know whether machine learning developments in particle physics are ahead or behind those in industry; for commercial reasons, the big data companies tend to be quite secretive about their technology. What we can say is that CERN’s open access approach to technology transfer has the potential to make machine learning techniques more widely available to smaller organisations.

So what are Adam’s plans when he completes his PhD next year? “Wherever I end up, whether it’s in academia or industry, I know I’ll be solving an interesting problem.”

 

Contents


 

Computer security is all about communication

Hannah Short

Hannah Short
(Credit: CERN)

 

We all know that computer security is important, but remembering numerous passwords is a drag. No wonder that more than 50% of people use the top 25 most common passwords (Source: Keeper).

You might think that the CERN community would be different, but it isn’t. “There’s a high likelihood that a typical university user will use trivial or repeated passwords,” says Hannah Short.

Hannah is a Fellow in the CERN Computer Security team and she’s working on an EC-funded project called AARC, which aims to make it easier for the academic and research community to collaborate by providing easy single-sign-on authentication and authorisation between institutes. The days of multiple passwords might be on the way out, but behind the scenes it isn’t so simple.

“This is the very human side of computer security,” says Hannah. “It’s all about building trust between institutes. There are protocols and framework, but if something goes wrong, you need to know the person, not the machine. If we don’t trust someone to run their research service securely, we’re not going to get very far!”

Hannah chairs WISE, a global community of internet security experts from different academic network and infrastructure providers. “WISE provides a forum to share our experiences and build trust – we can talk about common concerns and develop best practice.”

Hannah joined CERN in 2015. Although she had studied computational physics as part of her undergraduate degree, she hadn’t originally considered computing as a career. “The subject wasn’t offered at school, and I didn’t realise that computing was so enjoyable – programming is accessible, fun and you can earn money from it!”

That’s a message that Hannah is keen to pass on – she’s actively involved in outreach events aimed at encouraging more women and girls to consider a career in IT. “People think that jobs in computing don’t involve social interaction. Networking and relationship building are very important to my role. I’m bridging the gap between technology and people, but also between strategy and policy.”

Hannah’s combination of technical and communication skills gained community recognition recently in the form of a 2017 GEANT Award.

Contents


 

Data racks up at CERN

Cross bay infill

Cross bay infill
(Credit: Jeremy Hartley/Dataracks)

 

UK company Dataracks has supplied equipment to CERN’s new network hub.

The new building on CERN’s Prevessin site in France will ensure that there are no interruptions to the flow of data both internally and externally. Connected to the main Data Centre on CERN’s Meyrin site in Switzerland, the new network hub also has high speed fibre connections to the second data centre in Budapest as well as connections to all the key data locations across CERN.

Dataracks won the contract to supply 150 racks along with hot and cold aisle containment for high performance computing equipment following the company’s participation in one of the regular industry events arranged by the Department of International Trade and STFC. The equipment will be installed in the main Data Centre as well as the new network hub.

The UK@CERN events offer companies the chance to have one-to-one meetings with CERN project managers who are likely to be placing relevant contracts in the near future. The personal contact, and the opportunity to discuss the details of a project helps the companies to submit a successful bid when the tender is published.

Dataracks offers a bespoke service and Managing Director Jeremy Hartley believes that this has been a key part of the company’s success in winning the contract, “We offer a full service from design and manufacture to delivery and installation so we were able to work closely with the team in the IT department to get the specifications right and give them the precise functionality that they wanted from their racks.”

If you would like to add CERN to your client list, the first step is to make sure that you’re on the tender opportunities database.

The next UK@CERN event takes place on 27-29 November and places are still available. Contact Tender Opportunities for more information.

 

Contents


 

Inside story

Want to go inside the CERN Data Centre? On a racing drone? Of course you do!

On board a racing drone for a tour of the CERN Data Centre
(Credit: CERN)

 

Contents

Science and Technology Facilities Council Switchboard: 01793 442000