ASTRON & IBM Center for Exascale Technology

Newsletter 2

Dear reader,

In our newsletter from July 2013, we reported on the intermediate scientific results after approximately one year of research in the DOME project. In this edition, we would like to share some recent developments with you.

DOME is all about designing and creating new technologies for the Square Kilometre Array (the SKA radio telescope). Since July, there have been quite some developments on the SKA front. Last month, in October 2013, the SKA Office accepted the consortia´s biddings for the three-year design phase. This led to an initial planning and engineering meeting that same month and the kick-off of the pre-construction design phase in November. In terms of planning, this is a milestone, as it is now confirmed that the SKA-1 construction phase will start in 2017.

As DOME is a key factor in the Dutch preparations for SKA, it is now extending its reach and including a more intensive selection of the Dutch ICT academia, renowned worldwide for their  research in computer science. In this newsletter, you will read about a workshop  bringing together a number of top ICT professors and how they brainstormed about new avenues of research within project DOME.

But when it comes to a project as large and complex as SKA, we not only look for new science. We also try to learn as much as possible from previous experiences. Within DOME, we started this process early by analysing the insights gained from LOFAR and extrapolating our findings to the SKA context. This is one of the ways in which looking back helps us to shape SKA´s future.

Last but not least, we want to stimulate any positive effect that DOME research can have on Dutch society at large and on its high-tech industry in particular. Since the start of DOME in February 2012, several parties have joined the Users Platform. They are now collaborating with us on the development of new technologies for SKA. But we are getting ahead of ourselves: this is something you will read more about in our next newsletter.

We will keep you posted!

Albert-Jan Boonstra and Ton Engbersen

Scientific Directors DOME for ASTRON and IBM

 


 

 

SKA astronomers and engineers meet in Manchester

The Square Kilometre Array (SKA) takes not only a great amount of time to prepare, but also a great diversity of specialists. They generally work separately, making sure they share their expertise and insights on a regular basis. Last month, October 2013, a landmark SKA Engineering meeting was held in Manchester, the United Kingdom. Between the 7th and 11th October Michiel van Haarlem, Albert-Jan Boonstra and seventeen ASTRON colleagues joined as many as 230 project members in an impressive and highly informative team-building event.

Since 2011, separate teams of astronomers and engineers have been preparing for the next phase of the project. International consortia will take part in designing the elements of the SKA, with the office in Manchester as a central base, co-ordinating the preconstruction  activities.

Earlier this year the call for proposals was made, in order to begin the preliminary design and preconstruction phases of SKA. This will lead to key technical decisions being made, which will see SKA enter a new chapter. Agreeing and uniting the various teams behind the direction of the design was top of the agenda at the event.

Engineering meeting group at Manchester United, Old Trafford

Engineering meeting group at Manchester United, Old Trafford

 

 


 

 

World-class fundamental research on your doorstep

When seeking inspiration in scientific discovery we often automatically look beyond our national borders, expecting to find revolutionary insights and outstanding fundamental research in world-renowned institutes far away. But sometimes, we need not look very far to find a unique project of global significance. Take the quiet woods of Dwingeloo, for instance. Visionary work is being done right here, right now.

 

Artist’s impression of dishes South AfricaArtist’s impression of  sparse aperture array Australia
Artist’s impression of dishes (South Africa, left) and sparse aperture array (Australia, right)

 

Perhaps you or your organisation have already been playing a major role in this pioneering research? The Square Kilometre Array – better known as SKA – is being brought to life with the dedication and scientific knowledge of a large community of experts, including ASTRON and IBM. The two have joined forces to be able to meet SKA´s unprecedented demands in terms of data volume and computation speed. Their research will revolutionize the way we deal with big data.

 

A revolution in astronomy

SKA will be the largest and most sensitive radio telescope in the world. The continent-sized antennae array will be positioned in Australia and South Africa and will comprise of thousands of linked radio wave receptors. This will create an unrivalled radio wave collecting area, more sensitive and 10,000 times faster than any radio telescope array previously built. The aim is to observe a large volume of the universe, where human knowledge and understanding is still very limited. In the longer term, the results and findings generated from SKA will inevitably lead to new questions to be answered and advanced theories to be explored.

To continue its successful development, the SKA project relies heavily on the work of the associated national scientific research programmes and organisations. The benefits of industry collaborations ensure that SKA becomes a science and engineering partnership, spanning a range of technical and computing disciplines. However, the various research and development projects that combine to create SKA are both scientifically and internationally diverse. New technology and progress in fundamental engineering science are both required. Project DOME contributes to innovation on both ends.

 

SKA timeline

2003 Site selection begins
2012 Final selection of two locations: South Africa (dishes) and Australia (survey telescope and low-frequency array)
2013 Start of preconstruction phase
2014 Preliminary design
2015-2016 Detailed design phase and critical design review
2017 Start of SKA construction phase 1

 

 


 

 

Meet Dr. Michiel van Haarlem, Head of the Netherlands SKA Office

Michiel van HaarlemThe Square Kilometre Array (SKA) has been in the making for many years and preparations have gone through various stages. In 2011, Dr. Michiel van Haarlem was appointed Interim Director General of the Office of the SKA Organisation in Manchester. Under his reign, the SKA Program Development Office was replaced by the SKA Project Office. He led the transition from academic collaboration to the project-oriented organisation that is now overseeing the pre-construction phase. We spoke to Dr. Van Haarlem about the main challenges his organisation is now facing.

 

The road to Manchester

After receiving his PhD in astronomy from the University of Leiden in 1992, Michiel van Haarlem worked in industry for a brief period before taking up a post-doc position in Durham (UK). In 1996, he came to ASTRON where he worked on predictions of SKA observations. He became closely involved with the LOFAR project and was responsible for its design and construction as Managing Director. Dr. Van Haarlem moved to the SKA project in 2011, where he was appointed Interim Director General of the SKA Office in Manchester later that year. He returned to the Netherlands at the end of 2012, taking up the position of Head of the NL SKA Office.

 

LOFAR as pathfinder

Dr. Van Haarlem sets off by sharing some chronology with us: “It is interesting to note that the LOFAR telescope came out of the original planning for the SKA project, which was going on in the nineties.  SKA will have a large frequency coverage, whereas LOFAR focussed exclusively on the low-frequency end. This made LOFAR easier to implement technically. It also meant, that LOFAR equipment could be established and in use during SKA’s epic development process and design timeline. Indeed, important lessons are being learned from evaluating the functioning LOFAR equipment.”

As Van Haarlem underlines, “The existence of the low frequency antennae was critical to the successful development and continuation of the bigger and bolder SKA project. Researchers are able to build on the experience and knowledge gained from LOFAR.”

“LOFAR has become a bit of a champion. It has given us much more than a great basis for scientific pioneering. It has also maintained momentum and has fed the enthusiasm for all the preparatory SKA-work that is being done right now. There is active interest in SKA, not only within the astronomy community, but also far further afield.”

 

Challenges

Nevertheless, it is clear that SKA still faces many development challenges. The fact alone that the SKA antennae will produce an unprecedented amount of data, approximately ten times the current global Internet traffic every day, is daunting. It is much more than today’s supercomputers can handle. Van Haarlem explains, “The challenge will be to develop smart algorithms and high performance computing engines that are needed, which should also be energy-efficient. Transporting high data volumes across large distances will also be a key area of research.”

 

DOME & SKA

The research package Algorithms and Machines is DOME’s umbrella domain, steering and combining research and results from the other work streams. Its aim is to investigate the enormous complexities found in today’s radio telescope computing systems and machinery design, as well as lead to the development and use of a new analysis tool to benefit vast technical projects, such as SKA. The tool is a method directed at analysis and evaluation of the performance, power consumption, reliability and cost of a computing system. It will guide our decision making process when it comes to designing new computer architectures and algorithms.

“ASTRON’s work on SKA development is interwoven into the DOME work streams”, Van Haarlem says. “In the years to come, SKA phase two will aim to create an even more powerful telescope, building on the achievements and legacy of phase one.”

 


 

 

Top ICT professors gather around DOME

Wednesday 18 September 2013, one day after Prof. Dr. Michael A. Garrett received the IBM Faculty Award , a group of 22 people met in a conference room in Utrecht. Fifteen of the country’s top ICT professors were present, among whom Prof. Dr. Peter Apers and Prof. Dr. Arnold Smeulders, both members of the ICT roadmap team. Main purpose of this unique workshop was to identify common areas of interest, where current ICT topics and DOME-related research could be mutually reinforcing. A highly productive brainstorm ensued.

 

Interactive session

Host of the day was Dr. Robert van der Drift, Head of Computer Science at NWO, the Dutch Research Council. At the time of this workshop, NWO Physical Sciences was preparing a call for proposals, making funding available for new DOME-related ICT research. After Van der Drift set the scene, there were presentations by both of DOME’s scientific directors: Dr. Ton Engbersen (IBM Research Zurich) and Dr. Albert-Jan Boonstra (ASTRON). Trying very hard not to go into too much detail, they highlighted the main challenges for the SKA and each of the DOME research packages.

The session was markedly interactive. During the presentations, the professors put their questions or suggestions on the table right away. Discussions ranged from existing research that might be relevant to challenging research choices, fault tolerances, cost of materials, global networking requirements, migrating data and algorithms, statistical modeling, and more. And that was before the actual brainstorm had officially even started.

 

Top ICT professors gather around DOME

 

 

Professors with Post-its

The main event of the day was the brainstorm itself, during which participants were invited to share their first thoughts on possible research topics. They then wrote them down on Post-its and put them on the wall, each under one of the DOME research packages. The results were impressive: a host of potentially viable lines of research was now hanging on the wall.

It was Ton Engbersen’s job to scan them all, sometimes re-grouping them or changing the research package the ideas were attached to. In the end, 2 out of 8 DOME research packages were not linked to any proposals: ‘nano photonics’ and ‘software for microservers’ remained empty. The other six packages, however, generated a rich harvest of over 90 research topics.

 

Next steps

After the NWO call is issued in November, interested parties will receive further information and will be invited to make detailed yet compact research proposals. This call has the nature of a competition, with a panel of international experts reviewing all proposals. Key assessment criteria are scientific value, DOME-relevance, and knowledge utilization.  The same panel will then advise the board of NWO Physical Sciences on allocating funds. This will allow the ICT scientists involved to finalize assembling their research teams.

ICT already plays a major role in the DOME research project. By making additional funding available, the Dutch government, IBM and ASTRON are sending a signal which is crystal clear: innovation through and within ICT is essential for taking the next giant leap in radio astronomy.

 


 

 

What LOFAR teaches us about SKA

Rik JongeriusRik Jongerius is one of IBM’s talented young researchers, pursuing his PhD. Based at the ASTRON building in Dwingeloo, he works on DOME research package number 1: algorithms and machines. We spoke to him about the analytical model he has designed, applying the lessons learned from a large-scale Dutch telescope: LOFAR. He also tells us more about the search for high-performance yet low-energy systems needed for the Square Kilometre Array (SKA).

 

A model providing answers

One of Rik’s first goals was to design an analytical model, able to extrapolate LOFAR insights to SKA or exascale level. “That is a goal we have been able to meet. The model was completed in May 2013 and we have been testing and using it since then. It shows us where the bottlenecks in compute power and bandwidth are. We can change parameters like the number of antennas or the frequency band of interest, and it immediately calculates the new outcome. This is how we test various configurations and evaluate which is best for SKA.”

“We look at the whole process, starting with the antennas and ending with a picture of the sky. Our model is based on LOFAR, a pathfinder instrument for the SKA, with a processing pipeline that has proven to be effective. Therefore, LOFAR is a good place to start from. It operates on roughly the same principles as SKA will, only on a smaller data scale.”

 

As much as possible

Rik and his fellow DOME researchers are continually being challenged by various demands and interests. “On the one hand, it is essential to design a configuration which needs as little compute power and energy as possible. On the other hand there are the astronomers and what they will be able to do with the instrument. It is a matter of making possible as much as possible.”

What are some of Rik’s research questions? “Questions we ask ourselves on a daily basis are: how much energy will this take? Which hardware platform is the best match? How do we optimize performance, in such a way that it is very close to its theoretical maximum? Should we design our own chips? Is the initial investment worth the long-term gain in terms of energy efficiency, or should we use off-the-shelf components?”

 

New insights

But that is not all. “SKA will generate so many exaflops and petabytes, that it will need a system 100 to 1000 times more powerful than today’s most advanced supercomputers. So we may well need better algorithms to reduce the number of calculations. Or tinker with other parts of the system. Maybe we should start off with a smaller system for the first SKA science cases by the end of 2020 and expand from there. There are no final answers yet.”

Some new answers should come up soon, though. Rik Jongerius explains: “We have ordered some equipment that allows us to take energy measurements. I’m really looking forward to it, as it should yield some very interesting insights. Around February 2014, we should have some of the results.”

 


 

 

Prof. Dr. Michael Garrett wins IBM Faculty Award

Prof. Dr. Michael GarrettOn Tuesday 17 September, a select group of top researchers, professors and political leaders gathered at Leiden University, to witness the presentation of the IBM Faculty Award 2013. A special occasion, celebrating a decade of close collaboration between the Netherlands Institute for Radio Astronomy (ASTRON) and IBM,  the modest recipient, Prof. Dr. Michael A. Garrett, claims. We spoke to him about receiving the award, what it represents and what he plans to do with the endowment that comes with it.

 

Innovative algorithms

The ceremony started off with introductory speeches by Prof. Dr. Huub  Röttgering, the dean of Astrophysics at the University of Leiden who interrupted a trip abroad for the occasion, and Dr. Ton Engbersen, scientific director of the ASTRON & IBM Center for Exascale Technology. After that, it was IBM Europe’s chairman Harry van Dorenmalen MSc who did the honors and presented Prof. Garrett with the award.

Prof. Garrett is the general director of ASTRON and professor of radio astronomy techniques at the University of Leiden. He was given the award for research done by him and his team on algorithms which can be used to search for extraterrestrial life. This includes anomaly detection, statistical analysis and creating self-learning machines. All are highly innovative, multi-purpose and have great scientific and commercial potential.

 

Slightly embarrassing

“It has always been a team effort”, Garrett says, when asked what he did to win the IBM Faculty Award 2013. “I actually felt a bit embarrassed, but of course only one person can get it. The award symbolizes and recognizes great team work. One of ASTRON´s partners is IBM, there is a long history of collaboration. We’ve been working together for over a decade now.”

It is something special, this long-standing co-operation. “There aren’t that many collaborations between large commercial organizations and pure research organizations. When we started out, we set the scene for a much closer connection between radio astronomy and enterprise. It was one of the first major partnerships of its kind. No doubt, it has also encouraged other commercial entities and research organizations to join forces.”

 

Innovation through collaboration

Mike Garrett sees DOME as an essential step towards an inevitable and highly demanding future. “It’s not about building your own stuff and designing your own processes anymore, those days are long gone. In order to create the most advanced software, hardware, networking etcetera, you have to collaborate with the best. Under the DOME flag, that is exactly what we are doing.”

The IBM Faculty Award for Professor Garrett comes with an endowment of € 20,000.-. The money will go into SETI, the search for extraterrestrial intelligence. The chance that there is some form of intelligent life elsewhere in our universe is one of the reasons mankind has been probing the skies with big dishes since the nineteen sixties. It is also one of the drivers for innovation. “Whether or not we will ever find intelligent life on another planet, I don’t know. Whether or not we will even recognize the signal when it stares us in the face, I don’t know either. Is it worth the effort? That goes without saying.”

 


 

The Huffington Post ran an interesting piece on DOME on 13 November. Read the article.