Compression Methods for Simulation Models in SDM Systems

Models in Simulation Data Management (SDM) systems typically exhibit a great deal of redundancy. This is not being fully exploited by established compression techniques, such as zip. In view of the size of modern SDM systems, data storage and transfer cause large costs, which makes more advanced compression approaches necessary. Therefore, we consider two such advanced compression approaches: *Data deduplication* on the one hand exploits the mentioned redundancy, *mesh compression* on the other hand makes use of the specific mesh structure. The first approach, data deduplication, reduces space by removing repetitive data patterns. Every pattern is saved only once, and, wherever it reappears, it is replaced by a link to its first occurrence. We developed this approach into a viable procedure to be used in an SDM system; in the process we solved challenges such as choice of parameters, deletion of data, increased complexity of the data structure, exploiting the compression method for data transmission, and end-to-end encryption of the data for use by multiple parties. Using this procedure, we achieve compression ratios in the range of 3 to 8 for mixed SDM data; for pure simulation models we achieve 12 and more. The second approach, mesh compression, exploits the organisation of mesh data into vertices, which are connected by elements. For this we use a degree encoding algorithm, which traverses the mesh in a determined manner and stores the element type, valence of vertices or edges, and vertex data. Provided this information, it is possible to fully reconstruct the mesh. This approach is combined with geometry prediction, which improves the compression by transforming the vertices in suitable coordinates. In first applications quite promising compression ratios larger than 12 could be observed.