PURPOSE: To immediately detect a subsystem down with a low steady cost in a multiprocessor system.

CONSTITUTION: A down declaring means 12 which self-declares the down of its own subsystem 11 to a kernel at the time of the down of the subsystem 11, a down detecting means 14 which accepts, self-declaration of the down from the subsystem in each kernel, a down reporting means 15 which transmits a down report message to a preliminarily reported destination at the time of detecting the down by the down detecting means 14, and a down monitor means 16 which is operated on a certain processor and waits for the down report message from the down reporting means 15 to monitor the subsystem down are provided.