Currently, the efficiency improvement of complex technical systems is an urgent scientific task, which in numerous cases is formalized via optimization problem and solved by some well-known optimization methods. Metaheuristics are used intensively in this field due to their possibility to generate acceptable solutions in restricted time periods. However, it is known that with the decrease of performing time the solution accuracy of metaheuristics degrades, as well as algorithm’s convergence depends on its exploitation/exploration features and preliminary set parameters. Parallel metaheuristics implementations are used to improve algorithms performance. In this paper a distributed technique of computationally-hard optimization problems solving based on efficient workload assignment is presented and described. The novelty of the technique proposed is that the additional procedure of metaheuristics instances forming and distribution is added, which creates metaheuristic instances of a various computational complexity and assigns them to the most suited computing resources to perform parallel independent runs within the operation time restriction. The workload generated by metaheuristics blocks is distributed efficiently by means of metaheuristics portfolio usage. The latter makes it possible to improve the metaheuristics instances distribution and computational complexities forming so as to put the largest metaheuristic block to the node with the highest performance with the iterative improvement of the distribution, considering the full set of computing resources constraints and criteria.