Browse Prior Art Database

Disk Carrier Self Protection from Power Failure Write Data Corruption

IP.com Disclosure Number: IPCOM000112723D
Original Publication Date: 1994-Jun-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 86K

Publishing Venue

IBM

Related People

Morris, N: AUTHOR [+2]

Abstract

Disclosed is a method of monitoring input power at a device carrier to permit an orderly termination of write operations if power fails. Use is made of the "SCSI BUS RESET" to notify Early Power Off Warning.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Disk Carrier Self Protection from Power Failure Write Data Corruption

      Disclosed is a method of monitoring input power at a device
carrier to permit an orderly termination of write operations if power
fails.  Use is made of the "SCSI BUS RESET" to notify Early Power Off
Warning.

      Disk devices suffer data sector corruption if a power failure
occurs during a write operation.  Corruption creates a hard data
check in a sector on a fixed block device that normal error recovery
procedures cannot recover from which usually results in the sector
being reassigned and hence degrades the disk unit.  To prevent this
happening, Early Power Off Warning is required to enable the device
to terminate a write operation at a sector boundary and avoid data
corruption.  EPOW is usually provided by the power supply unit and
routed to a disk controller who then must attempt to cancel the write
operation before power is lost.  In many cases this is not possible.
Other implementations may cable the warning directly to the device
whose electronics would handle the warning.

      In high availability applications, multiple supplies are used
giving increased complexity to the warning eg a single failure
usually does not disturb the disk devices so a summing operation is
required to detect and warn of multiple failures.  This has meant
that even earlier warning is required from each to cope with
staggered warnings from the individual: e.g., 2mS warning from single
supply implementation requires 5mS from each unit in a cluster of 3
to guarantee the same 2mS from the last to collapse in the cluster.
The summing operation also requires additional logic to be located at
some convenient point and adds delay into the warning notification.
Increased cabling is required to route from the individual supplies
to the summing logic and onto the controller/disk devices.  This
disclosure is a simple cost effective method to supply EPOW to the
individual devices protecting them not only from power line
disturbances or power unit failure but also from power distribution
problems.

      In the implementation, disk devices are mounted on carriers
which accept input power from dual sources for high availability.

The sources are bulk DC voltages from mains input AC supplies each
carrier being connected...