Difference between revisions of "DSA Standard Disk Format"

From Computer History Wiki
Jump to: navigation, search
(New page)
 
m (Vaxorcist moved page DSDF to DSA Standard Disk Format: Just to create the redirect ...)
(No difference)

Revision as of 19:50, 20 August 2022

To present a perfect logical address space to the host and still permit interchange of drives and media among controllers, the DSA includes a physical media format standard called DSA Standard Disk Format (DSDF). DSDF divides a disk media physical address space into two subspaces, each identified by a unique header code:

  • The Logical block numbers (LBNs). This space, comprising the majority of the disk drive or pack, is the only one visible to the host. The majority of the LBNs are for user data storage. The remainder hold a multicopy, logically consistent record of known media defects called the replacement control table (RCT).
  • The replacement block numbers (RBNs). These blocks, physically intermixed with the LBNs with at least one on each track, are reserved for replacing LBNs which are defective at the time of manufacture or which become defective during use.

In addition, media for SDI drives contain two additional subspaces, also identified by unique header codes:

  • The external block numbers (XBNs). These contain a record of media defects at the time of manufacture, and are not normally overwritten in the field.
  • The diagnostic block numbers (DBNs). These are reserved for drive functional testing. It is possible to run SDI drive diagnostics without first copying user data to a backup medium, since diagnostics only write into DBNs, and not into the LBN user data area.

The quoted capacity of a DSA disk pack or drive (e.g. 456 MBytes for the RA81), is the formatted capacity consisting of the user data area of the LBN space. Defective or potentially defective blocks are removed from service and replaced by RBNs. After replacement, the user area remains the same size, and the user continues to address any replaced block by its original LBN. Host software is simplified and user capacity remains constant. Moreover, due to the extensive correction power of the ECC, replacement normally occurs before user data is lost.

The replacement process includes locating a free RBN as close as possible to the defective LBN, marking it used, storing the correct user data (or the result of attempted recovery) in it, and modifying the header of the original LBN to indicate the replacement. In UDA50 subsystems, replacement is performed with host assistance.

When it reads the header of a replaced block, the UDA50 revectors to the correct RBN transparently to the host, and uses it to store or retrieve user data.