ESCAPE project set to prepare for new era in supercomputing

30 April 2015

Efforts to prepare weather forecasting systems for the exascale era of supercomputing have received a boost after the European Commission backed the 4m-euro ESCAPE project, which is coordinated by ECMWF.

Earlier this year, the Commission favourably evaluated the proposal for ESCAPE submitted by ECMWF and 11 other participating companies and organizations under the EU’s Horizon 2020 Research and Innovation programme. A grant agreement is being drawn up and the project is scheduled to start on 1 October, subject to final confirmation of the grant by the Commission.

ESCAPE (Energy-efficient Scalable Algorithms for Weather Prediction at Exascale) will address the urgent need to adapt forecasting models to the massively parallel supercomputing facilities of the future.

Director of Forecasts Florence Rabier said: “ESCAPE is an important part of our drive to remain at the cutting edge of global medium-range weather prediction. We are delighted to be collaborating with other meteorological services, research institutes and hardware companies to bring this project to fruition.”

ECMWF’s partners in the project are Danmarks Meteorologiske Institut; Deutscher Wetterdienst; l'Institut Royal Météorologique de Belgique; Météo-France; MeteoSchweiz; Instytut Chemii Bioorganicznej Polskiej Akademii Nauk; Loughborough University; National University of Ireland, Galway; Bull SAS; NVIDIA Corporation; and Optalysys Ltd.

Peter Bauer, the head of ECMWF’s Scalability Programme, said: “This project will help us to continue to improve our forecasts while moving towards more energy-efficient, highly parallel supercomputers. At a time when climate change brings a growing risk of severe weather events and decision-makers increasingly rely on extended forecasting capabilities, our codes and algorithms must be adapted to work efficiently on tomorrow’s computer systems.”

The aim of ESCAPE is to develop world-class, extreme-scale computing capabilities for European operational numerical weather prediction (NWP). It will do this by defining fundamental algorithm building blocks to run the next generation of NWP on energy-efficient, heterogeneous high-performance computing (HPC) architectures. The project will pair world-leading NWP with innovative HPC solutions, fostering economic growth, EU business competitiveness and job creation.

NWP has been intimately connected with progress in supercomputing since the first numerical forecast was made about 65 years ago. But supercomputers are power-hungry and, with their current architectures, it will soon be impossible to deliver the required performance at a reasonable cost. Future, more energy-efficient systems with exascale capabilities (performing at least a billion billion calculations per second) will rely on parallel processing at levels to which current NWP codes are not adapted. There is therefore a need to make NWP processes and algorithms scalable so that they will work efficiently on tomorrow’s high-performance facilities.

For more information, visit our Scalability Programme page.

Optalysys Ltd has published a press release on ESCAPE.