Browse Prior Art Database

Reduction of Delay Experienced during Mirror Disk Access Failures

IP.com Disclosure Number: IPCOM000106837D
Original Publication Date: 1993-Dec-01
Included in the Prior Art Database: 2005-Mar-21
Document File: 2 page(s) / 85K

Publishing Venue

IBM

Related People

Shieh, JM: AUTHOR

Abstract

Disclosed is a method which will improve the recovery time experienced by a system using a mirrored disk configuration. Disk mirroring is a method by which a copy of one piece of information is kept at another physical location. With AIX, this is achieved by mirroring the logical volumes that are members of a volume group. On top of this logical volume, sits the Journaled File System (jfs). When a request is made to read or write to the jfs, the reads or writes are handed down from the jfs to the Logical Volume Manager's device driver (lvmdd). lvmdd in turn submits the read or write request to the disk device driver (diskdd). In a mirrored filesystem, when a write is required, the lvmdd issues two or more independent write requests to the diskdd. The one diskdd can control the actions of one or more physical disks.

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

Reduction of Delay Experienced during Mirror Disk Access Failures

      Disclosed is a method which will improve the recovery time
experienced by a system using a mirrored disk configuration.  Disk
mirroring is a method by which a copy of one piece of information is
kept at another physical location.  With AIX, this is achieved by
mirroring the logical volumes that are members of a volume group.  On
top of this logical volume, sits the Journaled File System (jfs).
When a request is made to read or write to the jfs, the reads or
writes are handed down from the jfs to the Logical Volume Manager's
device driver (lvmdd).  lvmdd in turn submits the read or write
request to the disk device driver (diskdd).  In a mirrored
filesystem, when a write is required, the lvmdd issues two or more
independent write requests to the diskdd.  The one diskdd can control
the actions of one or more physical disks.  When a read is required
from the mirrored filesystem, lvmdd issues only one read request from
the mirrors since all the data should be identical on the system.  In
a sequentially mirrored system, lvmdd issues the read to the primary
mirror.  If that primary request fails, then the lvmdd issues the
same request to the one or more disks that make up the secondaries.
In a parallel mirrored system, the lvmdd calculates which disks' head
location is closest to the read request and issues a read to that
disk.  Again, if the request fails lvmdd will issue another read to
the other disks that make up the mirrored filesystem.

      A problem arises when the user requires that the recovery from
a read or write failure to be very quick.  In the disk device driver,
each read or write request is given three chances of succeeding
before a failure is reported back to the calling program.  In the
case of mirroring, the diskdd is unaware of the fact that it is part
of a mirrored system and that there is another source of data that
can be accessed for reads.  For writes, if there is a failure, then
the data location can be relocated.  So, a retry count of three is
attempted during read or write problems that the disk may experience.
This retry causes a delay to the eventual, successful retrieval or
relocation of data.

      This proposal calls for an alternative method for handling
failures during data I/O. ...