We consider a preventive replacement policy for a cold-standby system with N components, in which only one component is in operation at a time. If the component in operation fails, a standby component is immediately switched into operation. If all components fail, the system fails. The system is inspected at random poins in time to determine whether it is to be replaced or not. If the number of failed components at the time of inspection exceeds a threshold value r, the system is replaced. Otherwise the decision is put off until the next inspection point arrives. Under the cost structure which includes a replacement cost, a system down-time cost and a holding cost of the components, we develop an efficient procedure to find the optimal control values N and r, which minimize the expected cost per unit time.