23 Mar 2017 Rome - At the Digital Day'17 event, hosted in Rome by the European Commission as part of the celebrations of the 60 years of the Treaties of Rome, Maria Chiara Carrozza, Member of the Italian Parliament and Professor at the University of Pisa, during the session about "Europe as a global player in High Performance Computing", moderated the panel that addressed the question of how Europe can regain leadership in HPC.
Xavier Prats Monné, Director General, Directorate General for Health and Food Safety (DG SANTE) at the European Commission, talked about High Performance Computing for biomedicine and specifically cited the example of rare diseases which poses a chronically disabling and even life threatening burden on some 30 million citizens in Europe. He explained that there are now 24 networks by different disciplines. Twenty-five countries are joining efforts to gain a critical mass for the treatment of rare diseases. Xavier Prats Monné claimed that this is an unprecedented example in Europe to exploit the potential of capacity. There needs to be a stronger capacity for diagnosis. Currently, it takes 5 years on average to define a diagnosis. This process must become much faster. Therefore, large investment is needed to break the fragmentation and the barrier between the different institutions.
Thierry Breton, Chairman and CEO of Atos SE in France, told the audience that the HPC session at the Digital Day event is an initiative that is extremely important for Europe. Atos is the only European company building large HPC machines at the moment. HPC requires a lot of know how. Atos has been building up that experience since 30 years. It are the industrialists who know how to build supercomputers. The European Commission has to have the skills to monitor what will be good and bad related to HPC. The industry can help the European Commission in that endeavour. Thierry Breton warned the audience against the emerging threat of protectionism which he thought was bad news. The US government are financing HPC and so is China and they both are protecting their business.
Moderator Maria Chiara Carrozza insisted on the important relationship between science and politics and asked Wolfgang Marquardt, Chairman of the Board of Directors at Forschungszentrum Jülich, Germany to share his viewpoint. He started by saying that his research institution does research on HPC but is also using HPC for science at the same time. The global race towards exascale will not be decided by the machine with the highest peak performance because a whole ecosystem is needed. Europe has always been a leader in HPC with an emphasis on software, algorithms, models and applications. There is a need to bridge the gap between methods and problem solving, Wolfgang Marquardt stated. Europe has a good reputation in several areas of scientific research, including astrophysics, material sciences, engineering, and finances. He claimed that the PRACE partnership is playing an important role in that it has a strong support of users groups from different disciplines. There are many examples of high-quality science but a different quality is emerging, which is the dissemination of information infrastructures. As an example of this Wolfgang Marquardt cited the Human Brain Project.
A weak spot in the European HPC landscape however is the development of processor technologies. The belief still exists that we can catch up and there is a fair chance, according to Wolfgang Marquardt. To achieve this, we have to join forces with technology providers in the mass market. The emerging disruptive technologies like quantum processors and neuromorphic processors also constitute an opportunity. However, European HPC should not be fully dependent on domestic micro-processors and electronics. There is a need for a dedicated balance between capacity and capability to establish the HPC ecosystem.
Scientific progress and economic growth are driven by HPC. In 2015 the US Department of Energy funded a study to shed light on the estimated profit of investment in HPC and this was even before artificial intelligence and Big Data were emerging. In the HPC domain, there are enormous expectations and possibilities but it is a question of combining the resources, human as well as financial resources, Wolfgang Marquardt concluded.
Susana Sargento, Associate Professor at the Universidade de Aveiro in Portugal, was representing the SMEs in the panel. She told the audience that today, we face lots of data. If, for example, you want to manage the city traffic, there is information on vehicles and on the people in the city and there is data from different sensors. The amount of data that needs to be stored and processed is enormous. This is the challenge that we have to face by deploying large Clouds and large HPC computing power. She stated that this is very important for many areas. Considering the growing world of Internet of Things, we need the information very fast and process the data in very few seconds or other wise, people might die. Therefore, there is a need to have everything digital and to be able to scale. A lot of computing power is required but we need to start from something that already exists, she told the audience. The biggest challenge of all is that we have to provide services for the citizens in order to solve the digital divide in the world.
Thomas Schulthess, Director at the Swiss National Supercomputing Centre (CSCS) in Lugano, Switzerland, said that the Human Brain Project has developed into a federated computing and data infrastructure. This has to be transparent for the user with an emphasis on the end user. Compute and data belong together, he said and he insisted that pragmatism without any bureaucratic burden is needed.
Europe is in the lead globally in climate modelling and midrange weather forecasting, he told the audience, and asked what the challenges were. One agreement in the community is that if we push the computation to the point where the models become simpler and reliable in the predictions, then we are on the right track. We need forecasts and studies on a longer time-scale. We need to push exascale computing because there is an economic value, Thomas Schulthess stated. As far as climate and weather forecasting are concerned, it is always good to know things sooner than later in order to mitigate the climate extremes. There is a need for a collaborative effort in Europe. Rather than focusing on exascale we need to focus on the scientific question, according to Thomas Schulthess. It is key to use the best technology that is available on the planet. He said that we are five or six years away from needing that technology and that the best integrators are needed. By being on top of the software technology we can be much faster than the US and China. Thomas Schulthess was convinced that Europe maintains its leadership on the scientific side.
Roberto Siagri, President and CEO of Eurotech SpA in Italy, wanted to put HPC into perspective. HPC was totally different in the past because the commoditisation of computers was not there yet. From 2000 to 2005 the consumer industry started to accelerate. Europe started to buy from other countries and lost its leadership there. Back then, we were living in a linear world, according to Roberto Siagri. Computing was just for scientists, and not industry. From 2010 to 2015, we discovered that Europe had to regain leadership and that it has to be in agreement with the US and Japan. Roberto Siagri stated that we should figure out how to close the gap before 2020. There is a lot of investment in the connectivity. Meanwhile Softbank has acquired ARM in the processor domain. Europe can close the gap if it buys the computers but that is not an option. Every delay in the delivery of components means that Europe will not be able to catch up with its competitors. If we look at the ranking, Europe is third, stated Roberto Siagri. The US has three times more capacity per capita. The big question is what will happen in five or ten years from now? Europe has to put more money into this and, most of all, Europe needs another dream for HPC, concluded Roberto Siagri.
Mateo Valero, Director at the Barcelona Supercomputing Center in Spain started by thanking everybody who believes in HPC in the European Commission. At the occasion of celebrating the 60 years of the Treaties of Rome, he mentioned that 70 years ago the transistor was invented. A little later, the DNA structure was discovered. Sixty years ago, also the research on artificial intelligence was started. With the development of the computer we were able to solve many problems. Spain was invited to join the European Union in 1986 and Spain was able, so to speak, to play the Champions League, according to Mateo Valero. In 1984 Barcelona was starting to use parallel computing. Barcelona was creating the school of computer architects. Barcelona established the first parallel computing centre in 1984. In 2005 the Barcelona Supercomputing Center was created where strong scaling hardware design and applications are being developed. Mateo Valero is trying to convince the European Commission not to loose out on hardware design. That is why the HPC Network of Excellence was created with more than 1000 European researchers.
Supercomputing centres in Europe are very good in applications, according to Mateo Valero, and not just in algorithm applications. The effort for hardware is necessary. If we do not produce hardware, we are not secure. We need to protect our companies, stated Mateo Valero. There is plenty of data to analyse DNA but the artificial intelligence field has been frozen until the researchers have good HPC power. He said that Europe will continue to play in second division if we do not have the right hardware. Europe likes to be independent but it needs to achieve this in collaboration with someone else. He concluded by saying that Europe can build processors and accelerators.
Moderator Maria Chiara Carrozza allowed Thierry Breton to have the final word. He said that there is a responsibility for using HPC for industrial leadership and asked whether it is possible to convince politicians and citizens. HPC is important for research but also for security. He insisted that we have to accept that supercomputers are important to protect the European continent and only Europe can provide that security. Therefore, we use HPC for defense. This is critical for the future, Thierry Breton concluded.