TRAFFIC REPLAY COMPRESSION (TRC): A HIGHLY EFFICIENT METHOD FOR HANDLING PARALLEL NUMERICAL SIMULATION DATA

Numerical simulations produce large amounts of data. A key challenge is how to cope with the resulting bottlenecks. This paper introduces a highly efficient data handling method using parallel inter-processor traffic, called TRC. It is used to reconstruct the internal fields inside one or more regions of interest (ROIs) defined by the user. The method is implemented on top of OpenFOAM®, an opensource computational fluid dynamics (CFD) platform [1]. A well known CFD benchmark is used to demonstrate its efficiency in terms of the space savings obtained. The benchmark test-case is validated using experimental data from the literature. Several compression algorithms are applied to assess the efficiency of the method within a user-defined ROI, including ZIP and RAR [2, 3]. It is shown that in terms of space savings TRC has an efficiency of an order of magnitude higher, and that the efficiency improves as the size of the uncompressed data gets bigger. The space savings obtained with classic algorithms remain constant for more than 60 Gb of floating point data. The obtained results bring new perspectives to attention regarding data handling techniques for parallel numerical simulations. Key words: numerical simulation, data handling, inter-processor communication, computational fluid dynamics, OpenFOAM