Towards green distributed storage systems

We model a distributed storage system consisting of energy harvesting nodes which store multiple files. To investigate the performance of file retrieval and node repair, we formulate two optimization problems: maximizing the number of retrieved (repaired) files given a deadline, and minimizing the retrieval (repair) time of a number of stored files. We derive the necessary and sufficient conditions for the feasibility of retrieving (repairing) a number of files by a deadline. Utilizing these conditions, we reduce the aforementioned problems to a single feasibility problem, which is solved using forward and backward algorithms. Finally, the system performance is illustrated numerically.