Dual subgradient method for constrained convex optimization problems