Non-recoverable read error projections
Let’s examine the expected number of non-recoverable read errors encountered when reading the entire contents of an HDD, as a function of the capacity of the HDD.
In Figure 1, the vertical axis is the mean number of non-recoverable errors encountered when reading the entire capacity of an HDD, and the horizontal axis is the capacity of the HDD in TB. (Some of you are thinking that this is now showing mean non-recoverable read errors, whereas the specifications state the maximum number. Since a maximum value is not statistically meaningful, and manufacturers don’t publish the shape of the distributions, I will stick with using it as the mean. If any manufacturers take issue with this approach, I urge them to publicly post their data so that all of us may use the proper values.) Figure 1 shows two lines, 1 for the consumer HDD specification of 1014 , and 1 for the enterprise capacity specification of 1015 .
One problem should be readily apparent – the non-recoverable read error specification isn’t scaling with capacity. 2TB disks are were introduced in 2009. At the consumer grade specification, there would be a 16% chance of encountering a non-recoverable read error when reading the entire capacity of such a disk. Enterprise-grade capacity disk isn’t too far behind – reaching about 4% on a 4TB disk. However unless HDD manufacturers make significant improvements in the non-recoverable read error rate, the situation will continue to degrade. At some point, the expectation value for a encountering a non-recoverable read error when reading the entire contents of a hard disk crosses 1.
We’ll look at the meaning of this in the next section.