Detection and Reallocation of Defective Clusters in Removable DASD Media
Original Publication Date: 1994-Jun-01
Included in the Prior Art Database: 2005-Mar-27
Arnold, AF: AUTHOR [+4]
Disclosed is a predictive sector reallocation technique for improving the data integrity of removable DASD (Direct Access Storage Device) media.
Detection and Reallocation of Defective Clusters in
a predictive sector reallocation technique for
improving the data integrity of removable DASD (Direct Access Storage
reliability of the medium, such as a removable magnetic
disk, of a DASD device degrades with time and use, due to the effects
of aging and due to the gradual development of "grown" defects. As
these effects progress, soft errors make it increasingly difficult to
retrieve data stored on the medium, and eventually hard errors make
the data completely inaccessible. When areas of the medium become
only marginally capable of storing data due to these effects, access
to the data can become intermittent. In the presence of intermittent
failures, it is particularly difficult to determine whether the
removable media or drive is at fault, so drives are frequently
replaced when the media is actually at fault, resulting in an
unnecessary expense and in a failure to fix the actual problem.
system typically attempts to read a sector five
times before a hard error is posted to the user. Laboratory data
indicates that, with flexible diskette media, the first retry is 92%
effective, the second retry is 2% effective, the third retry is 1%
effective, the fourth retry is 3% effective, and the fifth retry is
data recovered on the fourth or fifth retry is
"reallocated", being moved from marginal sectors to other sectors
before further aging renders the data irretrievable. The marginal
sectors are made unavailable for future data storage. Reallocation
may alternately be invoked on the second or third retry, providing
additional data security at the expense of making additional space on
the media unavailable due to an increased level of reallocation.
is implemented at the operating system level.
error types causing a sector to be considered for reallocation are
Sector Not Found, Data Cyclic Redundancy Check (CRC) Error, Data
Address Mark Not Found, and ID Address Mark Not Found. Data is
reallocated by the operating system on a cluster basis, with data in
a defective cluster being real located if more than three retries are
required to read the sector, up to the maximum number of retries
allowed by the operating system. The establishment of this threshold
level prevents unnecessary sector realloca...