Centralized versus distributed data