Browse Prior Art Database

Rebuild Only Used Space in Update-In-Place Disk Array

IP.com Disclosure Number: IPCOM000115788D
Original Publication Date: 1995-Jun-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 4 page(s) / 124K

Publishing Venue

IBM

Related People

Mattson, RL: AUTHOR [+2]

Abstract

It is common practice to use dedicated hot spare disks or dedicated spare space in disk arrays with parity so that when a disk fails, the data and/or parity information on the failed disk can be rebuilt from data on non-failed disks and then copied to empty spare disk space. Because the disk sub-system has no knowledge of what data lies in which blocks, every block on every disk belongs to some parity group and when a disk fails, data in every block on the failed disk is usually rebuilt and copied to spare disk space.

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

Rebuild Only Used Space in Update-In-Place Disk Array

      It is common practice to use dedicated hot spare disks or
dedicated spare space in disk arrays with parity so that when a disk
fails, the data and/or parity information on the failed disk can be
rebuilt from data on non-failed disks and then copied to empty spare
disk space.  Because the disk sub-system has no knowledge of what
data lies in which blocks, every block on every disk belongs to some
parity group and when a disk fails, data in every block on the failed
disk is usually rebuilt and copied to spare disk space.

      However, in most disk systems, the disks are not full of data ,
but rather, have space available for new Files, new DataSets, and/or
new Extents.  As a matter of fact, when the available data space is
nearly used up, the system usually signals the user to delete old
Files, DataSets, and/or Extents to make room for new data.

      If the disk array manager can deep track of which blocks
contain active data and which blocks contain available space, then
when a disk fails, only blocks containing active data need to be
rebuilt and copied to spare disk space which should speed up the
rebuild process.

      Update-In-Place (UIP) disk arrays with distributed parity and
distributed spare space usually have an organization such as shown in
Fig. 1a which illustrates the common notion dividing each disk into
segment-columns, containing N continuous blocks, and combining
corresponding segment-columns across disks in the array to define a
segment of N blocks by D disks for a total of N*D blocks/segment.  In
Fig. 1a there are N=5 blocks per segment-column and D=5 disks across
the array for a total of 25 blocks/segment.  Five segments are
illustrated in Fig. 1a, with disk blocks 0-4 in segment-1, disk
blocks 5-9 in segment-2, etc.  The symbols 3@, P@, and S@ are used to
indicate a data block on disk-3 in segment-@, a parity block in
segment-@ and a spare space block in segment-@ respectively.  Fig. 1
illustrates that if disk-3 fails, data and parity blocks on it can be
reconstructed from data and parity blocks on disks 1, 2, 4, and 5
shown in Fig. 1A, and that data copied into spare blocks such as
shown in Fig. 1b.  Since segment-column-3 on disk-3 contains spare
bl...