Failure prediction method for magnetic disk devices, and a magnetic disk device using the same

Abstract

A warning against the possible failure is accurately sent from the magnetic disk device to a host device. In one embodiment, error indices per unit time are stored discontinuously/continuously for each magnetic head, and if the new index (partial integral error rate) updated with the stored multiple error indices as the items to be monitored satisfies required conditions, this state is diagnosed as an abnormality or a pseudo-abnormality, and reported to the host device. The partial integral error rate is updated with an increase in the number of sets of error indices. Each time a new set of error indices per unit time are stored, the oldest of the existing error indices to be monitored is discarded and one of the new error indexes is added. In this way, a new partial integral error rate is calculated from the fixed number of new sets of error indices to be monitored. When the partial integral error rate increases above a second required threshold, this state is reported as an abnormality to the host device since a failure is most likely to occur in the magnetic disk device.