Abstract
This study considers the problem of optimal allocation of a common resource in systems with parallel structure. The distinguishing feature of suggested approach is its applicability to models complicated by randomness and possible, but not essential near singularity of the constraint matrix, which commonly arises in systems with parallel processes. The essence of the method is that it finds solution by first solving a simpler problem with an “expanded” constraint set, and then implements a directed transition to the solution of the original problem. In completing these steps the method takes account of the stochastic nature of the allocated flows.