Evolution of the Atlas data and computing model for a Tier-2 in the EGI infrastructure

ABSTRAC ISCG 2012 Evolution of the Atlas data and computing model for a Tier2 in the EGI infrastructure During last years the Atlas computing model has moved from a more strict design, where every Tier2 had a liaison and a network dependence from a Tier1, to a more meshed approach where every cloud could be connected. Evolution of ATLAS data models requires changes in ATLAS Tier2s policy for the data replication, dynamic data caching and remote data access. It also requires rethinking the network infrastructure to enable any Tier2 and associated Tier3 to easily connect to any Tier1 or Tier2. Tier2s are becoming more and more important in the ATLAS computing model as it allows more data to be readily accessible for analysis jobs to all users, independently of their geographical location. The Tier2s disk space has been reserved for real, simulated, calibration and alignment, group, and user data. A buffer disk space is needed for input and output data for simulations jobs. Tier2s are going to be used more efficiently. In this way Tier1s and Tier2s are becoming more equivalent for the network and the Hierarchy of Tier1, 2 is not longer so important. A number of concepts and challenges are raised in these proposals, and in this contribution we show how these changes affect an Atlas Tier2 and its co-located Tier3 that are using the EGI infrastructure. We will present the Tier2 and Tier3 facility setup, how do we get the data and the arrangements proposed to fulfil the requirements coming from the new model, like the fact that any site can replicate data from any other site. The approach to Dynamic Data Caching, where analysis sites receive datasets from any other site "on demand" based on usage pattern, and possibly using a dynamic placement of datasets by centrally managed replication of whole datasets, and unused data is removed. Also the Remote Data Access approach, where local jobs could access data stored at remote sites using local caching on a file or sub-file level. We also present how do we enable at the same time grid and local data access for our users, using the EGI infrastructure and procedures, and the middleware glite flavour that is being provided by EMI releases. In this direction an example of a real physics analysis and how the users are working will be presented, to check the readiness of the tools and how they perform with the current and within the changes being adopted coming from the evolution of the model