January 06, 2011
With petascale systems now deployed on three continents, the HPC industry is already looking toward the next milestone in supercomputing: exascale computing. In Europe, this activity is centered on the European Exascale Software Initiative (EESI), a project that brings together industry and government organizations committed to helping usher the transition from petascale to exascale systems over the next decade.
To learn more about the EESI, including the organization's activities and what took place at the recent workshop in Amsterdam, we spoke with two of the key players, including Jean-Yves Berthou, information technologies program director at EDF, and EESI program leader; and Peter Michielse, deputy director at NWO/NCF. We also got the US perspective from Jack Dongarra, who runs the Innovative Computing Laboratory at the University of Tennessee; and Pete Beckman, director of the Argonne Leadership Computing Facility.
HPCwire: Can you describe the EESI project for those who don't know what it is?
Jean-Yves Berthou: The European Exascale Software Initiative goal is to build a European vision and roadmap to address the challenge of performing scientific computing on the new generation of computers composed of millions of heterogeneous cores which will provide multi-petaflop performance in 2010 and exaflop performances in 2020. These hardware capabilities lead to outstanding technological breakthrough possibilities in computations and simulations, which will be reached only if an international cooperation work program is set up.
This is done through a set of conferences and work groups involving a very large number of HPC European actors, both scientific software developers and users. They will investigate where Europe stands in the overall international HPC landscape, what are its strengths and weaknesses, what are the priority actions, and what cooperation modes should be implemented between Europe and the international community. EESI will also identify the sources of competitiveness for Europe induced by the use of peta/exascale software. It will investigate and propose programs in education and training for the next generation of computational scientists.
The overall challenge must be faced at worldwide level to be attainable. EESI coordinates the European contribution to the International Exascale Software Project (IESP) launched by The Department Of Energy Office of Science and led by Jack Dongarra and Pete Beckman.
EESI is an FP7 Support Action funded by the European Commission under the call INFRA-2010-3.3: Coordination actions, conferences and studies supporting policy development, including international cooperation.
HPCwire: What are the timescales for the project?
Berthou: EESI has been launched on June 1, 2010 for an 18 months duration. A first mapping of the major HPC projects and organizations have be achieved. This mapping have been extended world-wide using IESP inputs and international contacts. This mapping is available on the EESI website.
The EESI workplan is now progressing in two directions. A first set of four working groups is targeting the technological computing domain challenges: hardware and associated software, computer science, numerical analysis and applicative software, that is, scientific and engineering codes. Each working group will produce its own roadmap by June 2011.
A second set of working groups will target the applicative side by looking for major grand challenges in climate and weather forecasting, industrial applications focusing on transportation and energy, physics and engineering sciences, and life science-health-BPM. Each working group will also produce its own roadmap integrating technological inputs identified by the first four working groups.
The economic dimension and impact on European competitiveness of these challenges will be specifically studied. To ensure close collaboration and sharing, one internal workshop will be held in February 2011, where each working group will be invited to present its results and roadmaps.
An overall synthesis will be produced and be presented at a large final public conference in Barcelona.
HPCwire: What is the funder, in this case, the European Commission, expecting to see as outputs from EESI?
Berthou: The expected outputs of the project is an exascale roadmap and set of recommendations to the funding agencies shared by the European HPC community, on software -- tools, methods and applications -- to be developed for this new generation of supercomputers.
HPCwire: Why does industry feel it is important to be involved in EESI?
Berthou: Exascale systems will engage the HPC community for the next 20 years in defining new generations of applications and simulation platforms. The challenge is particularly severe for multi-physics, multi-scale simulation platforms that will have to combine massively parallel software components developed independently from each others.
Another difficult issue is to deal with legacy codes, which are constantly evolving and have to stay in the forefront of their disciplines. This will require new numerical methods, code architectures, mesh generation tools, and visualization tools. In addition to the applications, all the software layers between the applications and the hardware need to be revisited for peta to exascale computers. Considering that 5 to 10 years are necessary to design, develop and validate a new generation of scientific applications, it is time now for industry to think about exaflop computing.
HPCwire: EESI recently held its first international workshop in Amsterdam. Can you tell us a little about that?
Peter Michielse: EESI held its first internal international workshop on November 9, 2010 in Amsterdam. The workshop has brought together approximately 80 -- mostly European -- experts in the areas of software development, performance analysis, applications knowledge, funding models and governance aspects in high performance computing.
An important part of the EESI project are the four working groups (WGs) in the area of application grand challenges and the four working groups in enabling techniques for exaflop computing. Each WG is composed of around 15 recognized experts, taking into account both expertise and geographical representation. The goal of each WG is to identify and classify the key challenges in their scientific area or technology component. This includes analysis of European strengths and weaknesses, existing collaborations, existing projects and opportunities for Europe.
During the morning session of the workshop, each WG presented itself, including the topics they view within their scope. Most WGs have been populated by experts, their first meetings have been planned, and an initial list of topics within each WG has been identified. During the discussion some aspects have been added to certain WGs.
The afternoon session started with an overview of the cartography results on HPC and exascale programs worldwide. It turns out that the DOE in the US is making progress in the areas of exascale software centers and co-design centers. Japan is developing its 10-plus petaflop K System, but along with that goes a strategic program on High Performance Computing Infrastructure (HPCI). On strategic programs in China, not so much is known. But it is a fact that developments and actual installations are taking place in petaflop systems that have put China on top of the TOP500 list. In addition, the European Commission, within FP7, has recently opened two calls with significant funding, dedicated to computing systems and exascale computing.
HPCwire: What were the main themes raised at the workshop?
Michielse: Basically, there were two important purposes for the meeting. First was to make sure that each WG was considering the right challenges within its scientific or technology field. During the presentations of the WGs, additional topics were recognized as being part of the WG, including several aspects which typically hold for more than one, or even for all WGs. These aspects include resilience, performance, power consumption and programmability of exascale software and systems.
The second purpose of the meeting was to get informed about US and Asian efforts with respect to their exascale software efforts, and as a result of that, investigate how the EESI Working Group activities align with those efforts and with the activities in IESP with respect to co-design of hardware, software and applications.
It also became clear that there are various challenges in international collaborations ahead, for instance synchronization of activities worldwide and organizational aspects to realize this, and also on how to cope with confidentiality of vendor developments and intellectual property rights.
HPCwire: What is the relationship between EESI and PRACE? And between EESI and other strategic activities in Europe, for example the recent IDC European HPC report? Are they competing or complementary?
Michielse: EESI, PRACE and other strategic initiatives in Europe are not only complementary, but should also strengthen each other. The IDC report gives its view on the opportunities for Europe with respect to future HPC, while the activities of PRACE are directed to building a pan-European infrastructure of Tier-0 HPC systems. The PRACE project not only investigates the actual infrastructure and regulations for that, but also heavily works on applications which are of high interest for European users and scientists.
An important role of EESI is to make sure that Europe is involved in global discussions on hardware, software and applications design and that Europe is involved in setting agendas and making choices for the benefit of European science, industry and economy. EESI could be viewed as the voice of the European HPC activities in a global context. Many people active in EESI are also active in PRACE and DEISA.
HPCwire: Supercomputing is often presented as a race, with nations vying for leadership to preserve industrial, economic and research competitiveness. How does the call for collaboration in exascale balance with this? Does this differ between hardware and software?
Jack Dongarra: Supercomputing capability benefits a broad range of industries, including energy, pharmaceutical, aircraft, automobile, entertainment, and others. More powerful computing capability will allow these diverse industries to more quickly engineer superior new products that could improve a nation's competitiveness. In addition, there are considerable flow-down benefits that will result from meeting both the hardware and software high performance computing challenges. These would include enhancements to smaller computer systems and many types of consumer electronics, from smartphones to cameras.
With respect to software, it seems clear that the scope of the effort to develop software for exascale must be truly international. In terms of its rationale, scientists in nearly every field now depend upon the software infrastructure of high-end computing to open up new areas of inquiry -- for example, the very small, very large, very hazardous, very complex -- to dramatically increase their research productivity, and to amplify the social and economic impact of their work.
It serves global scientific communities who need to work together on problems of global significance and leverage distributed resources in transnational configurations. In terms of feasibility, the dimensions of the task -- totally redesigning and recreating, in the period of just a few years, the massive software foundation of computational science in order to meet the new realities of extreme-scale computing -- are simply too large for any one country, or small consortium of countries, to undertake all on its own.
Standardization is also a minimum requirement for broad international collaboration on development of software components. In addition the international nature of the science will demand further development of global data management tools and standards for shared data.
Pete Beckman: One possible comparison to this effort is the International Space Station. With such a complex endeavor that targets scientific results that can benefit everyone, it is important to bring together collaborative teams of the best scientists from around the globe. By working together we can achieve more and deliver results sooner.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?