The objectives of this research are the theoretical development of efficient algorithms for large-scale optimization and the experimental examination of these algorithms. The approach is to use decomposition methods based on the dual decomposition technique. This technique takes advantage of additive separability which is often found in large-scale problems. The new algorithms will be applicable to both convex as well as non-convex optimization problems. In addition, methods will be developed for optimization of large-scale systems modeled by probabilistic discrete-event models.