Mechanical Relay (McR) is an architecture to enable data communications, where transmission delay and intermittent service are tolerated. Since each source and destination user uploads/downloads data when McRs are located in transmission range, we obtain more significant sum-throughput gain than fixed/no relay networks even if some packet delay might be burden for delay-sensitive cases. In this paper, we propose a couple of iterative algorithms: 1) water-filling power control (WFPC) for distance-based McR systems, 2) automatic initializing gradient algorithms (AIGA) that try to solve two optimization problems: 2-1) Sum-rate maximization, 2-2) Sum-MSE minimization. From the experimental results, utilizing McRs has much better sum-throughput performance than that of fixed/no relay cases, and WFBC also has the advantage in low SNR region. In addition, convergence points of general gradient methods highly depend on the step-size value, our proposed AIGA achieves lower complexity even if step-size is quite small.