Advancing Computational Fluid Dynamics with Grid Computing
The increase in computing power in recent years has helped unlock the potential of Computational Fluid Dynamics (CFD) in resolving longstanding fluid dynamics problems. As we are close to exhausting the upper limits of individual computers, a new paradigm has come to the fore that combines the computing power of several machines that are networked together. Grid Computing, as it is known, is today transforming the way research is performed, and CFD is no exception. The Information Society Technologies Programme funded six organisations to develop the necessary infrastructure to bring Grid Computing to CFD. The main deliverable of the project is the APUS-CFD solver, which employs numerical methods to solve the Navier-Stokes equations. Symban Power Systems Ltd, a FLOWGRID partner, ported an existing CFD solver into the Grid environment. Domain decomposition was key to enabling the CFD solver to operate efficiently in parallel mode. It was also necessary to separate communications tasks, which are handled with the Message Passage Interface MPICH-G2, from the mathematical computations. Additional developments included the creation of interfaces for easy modification of CFD components. Following the development stage, the APUS-CFD solver was tested by two FLOWGRID partners in Spain and Greece. The outcome was positive and established a precedent with Linux clusters as well as other configurations. Based on these encouraging results, the FLOWGRID consortium is pursuing commercial exploitation of the APUS-CFD solver. It will appeal to the automotive, aerospace and many other industries where CFD is an integral part of the product design phase.
