I will discuss the CERN experience in data handling to support the CERN physics experiments. All the physics data, notably the LHC data from Run 2, are stored in the CERN tape archival (120 PB), made available for the 150,000-core CERN computing farm and exported to the LHC Computing Grid (about 200 sites worldwide). The keystone of our system is EOS, the 140-PB CERN disk farm routinely delivering 30 GB/s or more. With the restart of the LHC programme, the data collected by the experiments exceed the 7 PB per month.

I will then discuss the evolution of the present system, especially in the direction of novel usage of the CERN data repository: data sharing via cloud storage (data synchronisation),  geographically-distributed data clusters for science and a closer integration between applications and data using Jupyter-notebook technology.