Browse Prior Art Database

Concurrent Customer Device Access during Media Maintenance

IP.com Disclosure Number: IPCOM000115666D
Original Publication Date: 1995-Jun-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 4 page(s) / 157K

Publishing Venue

IBM

Related People

Escola, DR: AUTHOR [+4]

Abstract

In Direct Access Storage (DAS) subsystems, Media Maintenance (MM) is the overall management process whose main goal is to locate and avoid defects that may cause read errors on the disk of a magnetic disk storage device. The MM function can be thought of as a number of key steps which we shall call MM operations. In previous DAS subsystems, the MM function was managed by system software by using I/O commands to the device. Most MM operations required many I/O commands which could involve a considerable amount of time. In addition, for those multi-I/O-command steps that required atomic operation, the system software was forced to use the Reserve I/O command.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 40% of the total text.

Concurrent Customer Device Access during Media Maintenance

      In Direct Access Storage (DAS) subsystems, Media Maintenance
(MM) is the overall management process whose main goal is to locate
and avoid defects that may cause read errors on the disk of a
magnetic disk storage device.  The MM function can be thought of as a
number of key steps which we shall call MM operations.  In previous
DAS subsystems, the MM function was managed by system software by
using I/O commands to the device.  Most MM operations required many
I/O commands which could involve a considerable amount of time.  In
addition, for those multi-I/O-command steps that required atomic
operation, the system software was forced to use the Reserve I/O
command.

This disclosure addresses the problem in 2 layers:
  1.  Moving the MM function within the subsystem
        Having the subsystem manage the MM function relieves the
         system software from performing many I/O operations to
perform
         1 MM step, and it allows the device to be accessed from
other
         systems while the MM function is taken through its steps.
  2.  Segmenting long-running MM operations
        Some of the MM operations require a large amount of time
         (greater than 1 second) to complete.  These operations would
         prevent host access to the device for an unacceptable time
         period.  Device I/O operations (from other systems) would
have
         a higher response time, and in some cases may even time out
         before completing their processing.  Segmenting long-running
MM
         operations allows concurrent use of the device by MM
operations
         and other I/O operations.

      The first step in allowing concurrent device access during MM
involved the movement of this function from the system level to the
subsystem level.  The subsystem is better able to manage the MM
function, and thus the subsystem can provide device access to others
in between steps of the MM function.

      The function also involves the segmentation of long-running MM
operations into sub-tasks.  The segmentation and associated
management is performed entirely within the subsystem.  The subsystem
determines if the MM operation is in the long-running category, and
if so, sets appropriate segment points.  At each segment point, the
task is queued at the tail of the device work queue, so the next MM
sub-task will be processed after all currently-queued device I/O
operations have completed.  In addition, only 1 MM operation is
allowed to be active at a given time, so new MM operations are
re-queued while a segmented MM operation is in progress.
  Note: This concept could be generalized for any category of
   commands: one could segment type-A commands ensuring only 1 active
   type-A command while sharing device access to all non-type-A
   commands.

      Each s...