Abstract
Recently distributed processing systems have been increasingly developed with various applications through advances in LSI technology and network technology.
In large scale distributed systems, the probability that some parts of the system are in failed state is high, moreover, the failed parts are repaired during operation of the system. Therefore, the system has come to require fault tolerance in order to keep its operation partially continueing by the cooperation of operational subsystems. For attaining this requirement, it is necessary for each subsystem to autonomously perform its function, that is, to administer its own resources and has its own self-contained fault detection and recovery mechanism.
In designing these systems described above, it is necessary to evaluate how to decompose the system in order to attain autonomously functioning subsystems. But this decomposition of the system has not been discussed and clarified.
In this paper, the method for evaluating the decomposition of system is proposed from the standpoint of fault tolerance. In this method, trade-off between loss and profit of the decomposition is evaluated by the functional reliability measure. Using this method, it is shown that the loss in the decomposition can be compensated by the profit in the decomposition. This method is applied to the loop network system and its effectiveness is shown by numerical examples.
