Browse Prior Art Database

Parity Preservation for Redundant Array of Independent Direct Access Storage Device Data Loss Minimization and Repair

IP.com Disclosure Number: IPCOM000104205D
Original Publication Date: 1993-Mar-01
Included in the Prior Art Database: 2005-Mar-18
Document File: 6 page(s) / 255K

Publishing Venue

IBM

Related People

Crews, CV: AUTHOR [+3]

Abstract

Described is a method to limit data loss in a Redundant Array of Independent DASD (RAID) and to provide a means to recover from the disruption of parity due to multiple errors in a RAID which protects against the loss of a single Direct Access Storage Device (DASD).

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 22% of the total text.

Parity Preservation for Redundant Array of Independent Direct Access Storage Device Data Loss Minimization and Repair

      Described is a method to limit data loss in a Redundant Array
of Independent DASD (RAID) and to provide a means to recover from the
disruption of parity due to multiple errors in a RAID which protects
against the loss of a single Direct Access Storage Device (DASD).

      Data storage devices can never guarantee complete protection
from data corruption or loss.  The most undesirable result is to
return bad data as if it were good data (i.e., with no error
indication).  Also very undesirable is that an error in a single
block causes the loss of a large amount of data.

      The purpose of this disclosure is to:

o   Eliminate data substitutions caused by an inability to update
    parity.

o   Prevent the loss of a single additional data block causing the
    loss of further multiple unrelated blocks.

o   Provide a means for recovering the use of a parity group having
    multiple failures among its constituent blocks.

The desired results are accomplished by the use of:

1.  A known data pattern (KDP) which replaces unknown user data when
    necessary.  The KDP may include extra non-user data such as a
    Longitudinal Redundancy Check (LRC), and must include as extra
    non-user data the bit described in the next bullet.  The LRC may
    or may not be inverted but if it is, it may provide an additional
    detection function as described below.

2.  A bit in the data record on the DASD but outside the user data
    (just as the LRC may be outside the user data) hereafter referred
    to as the corrupted data bit (CDB).  In a data record (not a
    parity record), the CDB on (bit=1) indicates that the user data
    is lost and the data returned is the KDP.  In a parity record,

    the CDB = 1 indicates that an odd number of data records in the
    corresponding parity group have the CDB = 1.

3.  A combination of hardware and/or microcode functions and/or
    algorithms which provide a means of maintaining the integrity of
    the parity group.  Physical reallocation of a data record is a
    necessary part of these algorithms.

      When a data record is encountered from which there is no
reasonable chance of recovering the data, and not recovering the data
jeopardizes the recovery of other data later and/or jeopardizes the
further use of the parity group, then the logical record is
reallocated (the LBA reassigned in SCSI terms) except that, if the
record is for a DASD known to be broken, this and the next step is
skipped, the reallocated data record is rewritten with the KDP, and
the parity record is regenerated using the KDP and data read from the
corresponding blocks of the remaining DASD of the parity group.  This
procedure is used recursively, if data from any of the remaining DASD
is not recovered, until a parity record is generate...