Event index-based analysis in the JUNO experiment

Purpose The Jiangmen Underground Neutrino Observatory (JUNO) is a large-scale neutrino experiment designed mainly for determining the neutrino mass hierarchy and accurately measuring the parameters of neutrino oscillation by detecting reactor antineutrinos. It is estimated that the detector will record approximately 5.2 TB of raw data every day, which only contain around 60 neutrino events. The time correlation analysis needs to select physics events from a very large data set.
Methods To address this challenge, an event index-based analysis software framework is designed and itsmajor developments include: 1) A new analysis workflow is implemented including pre-selection and event selection, and the former works as a data filter of the latter. 2) A new data format is defined to support reading indexed events efficiently for event pre-selection. In order to further improve the performance of data analysis, other developments are also completed, i.e. adding the support of multithreading based on TBB and moving computing-intensive components from data I/O service to the computation algorithm.
Results and conclusion The performance tests are completed on the simulated data which are close to real experimental data. The results show that reading data are 5 times faster from an IAD file than from a tree in the ROOT file. The multithreaded analysis exhibits high scalability. Performance studies also show that, with the event index method and multithreading, the efficiency of a typical analysis of JUNO can be improved by two orders of magnitude.