In LHC Run 3, the ALICE experiment will record 100 times more events than in the runs before. This is achieved with a continuous detector readout. To cope with such a huge amount of data, a new integrated Online-Offline (O2) computing infrastructure is created. Part of this development is a new analysis framework.
In Run 1 and Run 2 a large fraction of the time to analyze a dataset has been used to read the data from disk and to deserialize it. Consequently a main focus of the new development is on this step. A new flat data structure is developed which supports growing, skimming and pruning of the data. Only the absolutely necessary information for an analysis will be read. Uninteresting event or track information should not be processed. The important parts of the dataset will be read at dedicated analysis facilities which are themselves optimized for a quick data access.
The new framework should be built on open source implementations to reduce the required work within ALICE and to benefit from synergies with industry. By using open source implementations, new code can be contributed and no dependency on proprietary software is created.