Although this issue seems to be more of concern to the efficient design of an application that runs on the grid, it would be preferable if such storage management functionality would
be offered directly by the grid middle ware, and handled seamlessly by merely replacing the original data with the processed data. If the processing application is run as a service this would merely require invoking the desired service while pointing to the data stored on the grid, which would be than automatically replaced by the processing results upon the successful completion of the grid job. This scenario would completely eliminate the phase 7 of the grid job submission.