Impact of ROOT file parameters on ATLAS Analysis Object Data

ATLAS Analysis Object Data (xAOD) and derivative representations (DxAOD) are based on the ROOT file format. In this study, we investigate the impact of ROOT file parameters on file size and reading speeds of (D)xAODs in common ATLAS workflows. This covers LZMA, ZLIB and LZ4 compression at different levels, as well as ROOT features like Autoflush and the TTreeCache. Recommendations are given for different scenarios, mainly with respect to different event sizes. These recommendations can be used as a starting point for detailed studies of new usage scenarios or new proposed file formats.