Randomized methods have a long history of applications in control and optimization problems. We review some classical methods of random search and random approach to robustness, going back to Rastrigin, Stengel and Ray and many others. Monte Carlo technique is the basis for many such methods. New versions of this technique (so called MCMC – Markov Chain Monte-Carlo) became popular recently. We survey some of these methods (including Hit-and-Run and Shake-and-Bake) and provide more efficient versions, based on exploiting barrier methods of convex optimization.