The first exascale supercomputers are expected to arrive at the end of this decade. These systems must deliver an exaflop - or a million million million calculations per second. A new consortium of leading European sites has come together to meet the challenge of delivering software to support and exploit these massive systems.
The 8.3M Euro CRESTA project - Collaborative Research into Exascale Systemware, Tools and Applications - brings together thirteen leading European sites to deliver this technology. The project is funded under the EU Framework 7 research programme and represents part of a significant investment in exascale research by Europe.
The CRESTA consortium has shown that software is critical to successfully achieving exascale performance.
As Dr Mark Parsons, the project coordinator for CRESTA and the Executive Director of EPCC at the University of Edinburgh, points out, "The primary challenge of exascale computing is developing software for these enormous systems. The CRESTA consortium brings together a unique group of experts in this field."
CRESTA's experts include four of Europe's leading HPC centres (EPCC, HLRS, CSC and PDC), a world-leading supplier of HPC systems (Cray Europe), seven application owners from science and industry (DLR, KTH, ABO, JYU, UCL, ECMWF and CRSA), Europe's leading HPC tool company (Allinea) and Europe's leading performance analysis organisation (TUD).
The importance of this integrated approach is explained by Dr Erwin Laure, director of PDC's High Performance Computing Centre: "Exascale computing is posing new challenges on the entire software stack. A multidisciplinary approach combining hardware specialists, computer scientists, mathematicians, application developers and domain scientists is needed to address these challenges. CRESTA brings all of these competences together in co-design teams working on important European software packages to foster Europe's leadership in HPC software."
The project has two integrated strands linked via a cyclical co-design process: one focused on enabling a key set of co-design applications for exascale, the other focused on building and exploring appropriate "systemware" for exascale platforms.
Dr.-Ing. Stefan Wesner, Managing Director at the High Performance Computing Centre Stuttgart (HLRS), explains: "The Exascale computing challenge cannot be solved with existing approaches. Thinking in layers where hardware platform, operating system, tools, libraries and applications are optimized independently from each other is insufficient. Radical changes would be limited to a single aspect. The co-design process implemented in CRESTA targeting joint application and systemware development is the right answer for addressing this
The six co-design applications represent an exceptional group of applications used by European academia and industry to solve critical grand challenge issues, including: biomolecular systems, fusion energy,the virtual physiological human, numerical weather prediction and engineering. The project also brings together some of Europe~Rs best tools developers to examine issues around debugging, performance analysis and pre- and post-processing.
The project's focus on these areas is unique. David Lecomber, CTO of Allinea Software, the company behind the petascale debugger Allinea DDT, makes the case: "In addressing the needs of the scientific software for Exascale, CRESTA provides a vital focus for real users with real applications and tool requirements."
More about the project and the partners can be found at: http://www.cresta-project.eu
UCL breaks through scaling roadblock with development tools at scale and achieves first si…
CRESTA’s remit involves preparing applications and software for future exascale platforms,…