Data Management for the LHC Experiments: Requirements and Challenges that drive the gLite Middleware Development

Data management has been from the start on the most critical aspect of grid computing for the High Energy Physics community. The obvious challenges are posed by the multi petabyte scale of the required storage, the different access patterns and of the several thousand users that need dependable, high bandwidth access to the data. The less obvious, but as much challenging and critical requirements are rooted in the need to operate a highly distributed, loosely coupled, set of facilities in a coordinated way. The current gLite data management software will be presented, an overview of the operational experience will be given and some of the reasoning that led to specific technical choices will be discussed. Based on the lessons learned in almost 10 years of grid data management, an outlook on the possible future evolution of grid data management will be given.