In this paper, we show how convex optimization in combination with a suitable parametrization of all stable closed-loop systems can be used to determine the best possible performance for single-input single-output, multi-input multi-output and decentralized controllers. The approach can handle most practically relevant design specifications, no translation of the original specification into weighting functions etc. is required. This constitutes an effective tool to explore different control structures and to integrate the performance of the controlled plant into the plant optimization process.