Every second year the European Centre for Medium-Range Weather Forecasts (ECMWF) hosts a workshop on the use of high performance computing in meteorology. The 14th ECMWF Workshop on High Performance Computing in Meteorology was held from 1 to 5 November 2010.
Description
The emphasis of this workshop was on running meteorological applications at sustained teraflops performance in a production environment. Particular emphasis was placed on the future scalability of NWP codes and the tools and development environments to facilitate this.
The aim was to provide a venue where:
- Users from our Member States and around the world could report on their experience and achievements in the field of high performance computing during the last two years; plans for the future and requirements for computing power were also presented.
- Vendors of supercomputers had the opportunity to talk to managers and end users of meteorological computer centres about their current and future products.
- Meteorological scientists could present their achievements in the development of parallel computing techniques and algorithms, and could exchange ideas on the use of supercomputers in future research.
- Computer scientists could give an update on their efforts in providing tools which will help users to exploit the power of supercomputers in the field of meteorology.
- The challenges of creating a computer centre infrastructure for HPC could be discussed.
The workshop consisted of a limited number of presentations from invited speakers, plus a series of 20 minute contributions. The morning of the final day was reserved for an open discussion session.
Programme
Presentations
Session 1 |
|
|
Development of the ECMWF forecasting system Erland Kallen (ECMWF) |
 |
|
IFS scalability and computational efficiency Deborah Salmond and Mats Hamrud (ECMWF) |
|
Session 2 |
|
|
Recent advancements in high resolution climate modeling Richard Loft (NCAR) |
|
|
The next-generation supercomputer and NWP system of JMA Masami Narita (Japan Meteorological Agency) |
|
|
HPC at the Canadian Meteorological Centre Bertrand Denis (Environment Canada) |
|
Session 3 |
|
|
4D-Var, scalability and code design Yannick Tremolet (ECMWF) |
|
|
A validation of the Yin-Yang global forecast GEM model Vivian Lee (Environment Canada) |
|
|
US navy mesoscale forecast system and HPC attributes James Doyle (Naval Research Laboratory) |
|
Session 4 |
|
|
Using GPUs to run weather prediction models Mark Govett (NOAA Earth System Research Laboratory) |
|
|
Progress on the GPU parallelization and optimization of the NIM global weather model Tom Henderson (NOAA Eath System Research Laboratory) |
|
Session 5 |
|
|
Petascale opportunities and challenges for Earth System Modeling Per Nyberg (Cray Inc) |
|
|
Application scaling in an Exascale Environment Don Grice (IBM Corporation) |
|
Session 6 |
|
|
HPC in the multi-core and heterogeneous era David Barkai (Intel Corporation) |
|
|
Fujitsu's approach for application centric Petascale computing Okuda Motoi (Fujitsu Limited) |
|
|
Update on SGI technology for weather and climate simulation Michael Woodacre (SGI) |
|
Session 7 |
|
|
Performance "Barriers" in highly scaling earth-system models Panagiotis Adamidis (Deutsches Klimarechenzentrum GmbH) |
|
|
High performance computing at CMA Xiao Huadong (China Meteorological Administration) |
|
|
An Architecture Comparison Exercise (ACE) Lois Steenman-Clark (University of Reading) |
|
Session 8 |
|
|
High performance computing of MSSG and its physical performance Keiko Takahashi (Earth Simulator Center, JAMSTEC) |
|
|
Future activities of high performance computing in meteorology using the K-computer Hirofumi Tomita (Japan Agency for Marine-Earth Science and Technology) |
|
|
IFS: RAPS11 and model scaling George Mozdzynski (ECMWF) |
|
Session 9 |
|
|
RAPS update George Mozdzynski (ECMWF) |
|
|
Future trends in high performance computing at DWD Ulrich Schattler (Deutscher Wetterdienst) |
|
|
Meteo-France RAPS benchmark Phillipe Marguinaud (Meteo France) |
|
Session 10 |
|
|
The challenges of running meteorological applications at sustained teraflops performance with resilience and consistency in a production environment Tim Pugh (Australian Bureau of Meteorology) |
|
|
Scalability of the Met Office unified model Andy Malcolm (Met Office) |
|
|
Operational numerical weather prediction in Switzerland and evolution towards new supercomputer architectures Philippe Steiner (Meteo Swiss), Michele de Lorenzi, Angelo Mangili (CSCS) |
|
Session 11 |
|
|
Bull's high performance computing solutions for science simulations Jean-Pierre Panziera (Bull) |
|
|
Accelerating weather models with GPGPU's Dave Norton (The Portland Group) |
|
|
Fortran 2008: what's in it for high-performance computing? John Reid (Rutherford Appleton Laboratory) |
|
Session 12 |
|
|
Parallel I/O for scalable earth system modelling Luis Kornblueh (Max Planck Institute for Meteorology) |
|
|
HPC weather and climate code development with Rogue Wave Dean Stewart (Rogue Wave Software UK Ltd) |
|
|
Debugging at scale Lindon Locks (Allinea Software) |
|
Session 13 |
|
|
Enabling exascale computing through the ParalleX execution model Thomas Sterling (Louisiana State University) |
|
|
Ensemble data assimilation as parallel minimization Tuomo Kauranne (Lappeenranta University of Technology) |
|
Session 14 |
|
|
Parallel processing of the semi-lagrangian scheme in GRAPES ZhiYan Jin (China Meteorological Administration) |
|
|
ODB (Observational DataBase): past, present and future Anne Fouilloux (ECMWF) |
|
|
Preliminary results of GRAPES' Helmholtz solver using GCR and PETSc tools XiangJun Wu (China Meteorological Administration) |
|
Session 15 |
|
|
Efficient local resorting and preconditioning techniques applied to a parallel tsunami simulation model Annika Fuchs (Alfred Wegener Institute for Polar & Marine Research) |
|
|
The NCEP HPC workload: 2010 state and forward challenges George VandenBerghe (NOAA/NCEP) |
|
|
I/O monitoring of 4D-Var at ECMWF John Hague (IBM) and Oliver Treiber (ECMWF) |
|
Session 16 |
|
|
GLOMAP mode on HECToR phase2b (Cray XT6) Mark Richardson (Numerical Algorithms Group) |
|
|
The data processing concept and performance requirements for the Meteosat Third Generation Programme Dieter Just (EUMETSAT) |
|
Session 17 |
|
|
Optimizing a vector code for cache-based multi-core systems: Porting echam6hms Luis Kornblueh (Max Planck Institute for Meteorology) |
|
|
A cost-effective redundancy scheme for real-time data production systems through the usage of virtualization Martin Dillman (EUMETSAT) |
|
Session 18 |
|
| Panel on experience of using High Performance Computing in Meteorology | |