A multicomputer system is considered in which the fault-and-failure to
lerance is established through the solution of the some problem by a f
ew computers that form a complex with the subsequent interchange of th
e results for the choice of a correct solution. It is assumed that the
system consists of two completes of this type to solve in parallel di
fferent problems which involve the interchange of messages over interc
omplex: communication channels. A method is suggested for the accompli
shment of the system operation in such a way that the system carries o
ut the continuous, ripple-through, and mutually registerable detection
and identification of intracomplex and intercomplex symptoms of singl
e ''hostile'' malfunctions, so that the system makes a simultaneous an
d complex-consistent decision in the event of the identification of a
failure and the localization of the site of its origin.