In numerical weather and climate prediction (NWP), accuracy depends strongly on available computing power ? but the increasing number of cores in top systems is leading to a higher frequency of hardware and software failures for NWP simulations. This report (from researchers at eight different institutions) examines approaches for fault tolerance in numerical algorithms and system resilience in parallel simulations for those NWP tools.
Authors: Tommaso Benacchio, Luca Bonaventura, Mirco Altenbernd, Chris D. Cantwell, Peter D. Düben, Mike Gillard, Luc Giraud, Dominik Göddeke, Erwan Raffin, Keita Teranishi and Nils Wedi.