Choose timezone
Your profile timezone:
The LHC experiments described the data replication policy in theircomputing models, already in 2006. The initial models were very conservative and foresaw a very high replication factor for the processed data types available to the physicists for analysis and a flat data distribution over the T1 and T2 centers. Experience however has shown that only a very limited subset of data is heavily accessed and a flat data distribution model creates unnecessary load on Grid resources such as storage and network, beside occupying a lot of disk space. Various experiments therefore decided to implement a more dynamic model where data replication and subsequent data cleaning is partially based on the real data usage. In this contribution we will describe the dynamic data placement system implemented by ATLAS and the related work in the CMS experiment.