A system is considered in which two dual nodes with databases are placed at
distant locations for mutual backup. When they are distributed in the netw
ork, the relationship between the centralized maintenance and the reliabili
ty is studied. The maintenance teams are centrally located at the maintenan
ce center. The maintenance is carried out by two types of calling, emergenc
y calling, requiring response in a short time, and nonemergency calling, wi
th response after a long wait, depending on the system failure conditions.
In recognition of temporal reliability, the degree of emergency is classifi
ed into several levels depending on the number of failed devices in the sys
tem and on whether the maintenance team has arrived. Based on this level, t
he method for maintenance is determined. Using this proposal, the system re
liability and the number of callings are systematically evaluated. For none
mergency calling, several failures are scheduled, while for emergency calli
ng the number is as small as possible, so that the maintenance method is ma
de efficient. (C) 1999 Scripta Technica, Electron Comm Jpn.