Experiments and simulations produce large volumes of data stored on large tape systems (known as HPSS) and distributed disk systems. The dCache system is a management system for a number of high energy physics experiments, such as ATLAS. It maintains disk spaces for permanent files and temporary disk caches for HPSS connectivity. Since this dCache installation is shared by a number of physics research groups around the country, it is challenging to balance the usage demands from all the research groups. In this work, we study the popularity of the data sets used by the research groups with the aim to determine the predictability of these usage patterns. We are working with operators of the dCache system at for ATLAS experiment to ensure the patterns are grounded in real application usage. Ultimately, we want to help the ATLAS collaboration to reduce their data access time.
