Optimization of the read performance of ROOT files at LHCb

In this qualtitative study an arbitrary physics analysis of the LHCb experiment is simulated. In such an analysis one wants to read files which contain the information from particle collisions produced by the Large Hadron Collider.\\ In this report an attempt to assess and optimize the reading of these datafiles is done. Clearly a fast execution is favoured.\\ The ROOT I/O mechanism is the main mechanism to populate the Transient Event Store. This mechanism provides a number of parameters to optimize the read performance.\\ The influence of these parameters when reading the data locally or remotely is investigated and effects resulting from changing the access patterns are studied.\\ During this analysis it becomes clear that the ROOT TTree cache size should be between 10 and 20 MB when reading LHCb data. Cache sizes bigger than 20 MB do not have further impact on the mean load time of the accessed branches for the data produced in an average LHCb analysis data set. Furthermore "data reclustering" becomes important. To optimize the read performance one should put all the interesting events that are read often in a file to minimize "holes" of insignificant data.