An algorithm was proposed to extract, if possible, from an arbitrary-structure network an environment of mutual informational system agreement including the problem solution complexes and the environment of intercomplex exchange. Such agreement is required in the computer network to organize parallel solution of interacting problems with given characteristics of their fault-tolerance based on the dynamic redundance. The proposed approach, methods, and algorithms are applicable to the distributed systems, as well as the grid and "cloud" computations for organization of computations with the desired confidence.