This paper presents joint design of source power allocation and relay beamforming for multi-user multi-relay wireless networks. The problem is formulated as maximization of the worst signal-to-inference-plus-noise ratio (SINR) among users subject to various power constraints. An equivalent d.c. (difference of two convex functions) programming based method is then used to develop an iterative algorithm with low computational complexity to obtain the optimal solution. Simulation results demonstrate that our optimal solutions of the joint optimization problem significantly outperform the solution obtained by separate designs of source power and relay beamforming.