It is shown an incorrectness of introduction of a class of NP-complete problems, which reason is that Cook s S.A. theorem on that the satisfiability problem is the universal NP-complete problem, is not true and, therefore, the issue on existence of at least one NP-complete problem remains open, that explains failures of attempts to estimate correlations between P and NP classes. Developed a universal procedure for solving a wide range of discrete optimization problems, which can be efficiently parallelized based on cyclic structures.