This paper discusses generalized methods on the optimal 1/N backup procedure of data to a floppy disk. First, assuming that the failure time of a data floppy disk follows a common probability distribution, necessary and sufficient conditions on the existence of the optimal 1/N backup policy which minimizes a relevant expected cost function are analytically derived. Second, we propose two nonparametric methods to estimate directly the optimal backup policy from failure data, when the failure time distribution is unknown. Finally, asymptotic properties for the estimator of the optimal backup policy are compared, and the relationship between the number of failure data and the precision of the estimate is examined through numerical examples.