ACCOLADES: A Scalable Workflow Framework for Large-Scale Simulation and Analyses of Automotive Engines

Analysis and optimization of simulation-generated data have myriads of scientific and industrial applications. Fuel consumption and emissions over the entire drive cycle of a large fleet of vehicles is an example of such an application and the focus of this study. Temporal variation of fuel consumption and emissions in an automotive engine are functions of over twenty variables. Determining relationships between fuel consumption or emissions and the dependent variables plays a crucial role in designing an automotive engine. This paper describes the development of ACCOLADES (Advanced Concurrent COmputing for LArge-scale Dynamic Engine Simulations), a scalable workflow framework that exploits the task parallelism inherent in such analyses by using large-scale computing. Excellent weak scaling is observed on 4,096 cores of both an Intel Sandy Bridge-based cluster and a Blue-Gene/Q supercomputer.