BAD SECTORS

dec_rd54a_bad_sector_map_700In the early days of hard disks, manufacturers placed bad sector tables on the labels of the disks. This was done so that manual marking of bad sectors could be done.

The image shows an old MFM  hard disk with its defect table. This was typical of early hard disks.

MS-DOS formats disks completely and then scans each sector to be sure it is readable. Any unreadable sectors are then marked off in the file allocation table.

Unfortunately some sectors are readable but they are not working as good as they should. The DOS PC Tools was very popular with its disk checker that could more reliably identify defects.

With the advent of EIDE disks and integrated controllers the industry started using more and more abstractions. By using sector spares and address translation the disk eventually became a block storage device with some specified number of storage spaces.

Bad sectors may manifest some time after the disk is installed. Marginal areas on the media may become unreliable and they must be identified and reallocated.

More recent disks have used advanced error correction to reduce the hard error rates. Early recovery started at 1012 before rising with capacity growth. Years later when 4096 byte sectors became necessary, the hard error improved to 1014.

EIDE disks may show bad blocks. This happens when there are more bad blocks than recovery blocks. Hard disks are designed to minimize the emergence of bad blocks.

Some consider bad block to be a sign of impending disk failure. We have used many disks with back block for over 10,000 operating hours successfully which is a counterexample.

DATA SCRUBBING

One way to keep data safe is to use data scrubbing. My using different checksums it’s possible to identify damaged files and then it can be recovered.

ZFS was designed for storage and it has verification against data corruption modes, continuous integrity checking, and automatic repair.  FreeNAS using ZFS with multiple disks to provide a more fault tolerant storage solution. ZFS does not work with RAID so cards should use JBOD mode. ZFS does need more RAM installed to cope with deduplication and copy on write with larger capacity disks.