Browse Prior Art Database

Limited Distributed DASD Checksum, a RAID Hybrid

IP.com Disclosure Number: IPCOM000109640D
Original Publication Date: 1992-Sep-01
Included in the Prior Art Database: 2005-Mar-24
Document File: 2 page(s) / 117K

Publishing Venue

IBM

Related People

Bashore, TJ: AUTHOR [+3]

Abstract

A Redundant Array of Inexpensive Disk (RAID) that fully distributes parity across all arms (RAID-5) performs better than a RAID that has a single parity drive (RAID-4). But with RAID-5, the usable capacity (viewed by the system) of all disk units changes when units are added or removed from the checksum set. This adds to the complexity of attaching a RAID-5 box to an existing system.

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

Limited Distributed DASD Checksum, a RAID Hybrid

       A Redundant Array of Inexpensive Disk (RAID) that fully
distributes parity across all arms (RAID-5) performs better than a
RAID that has a single parity drive (RAID-4).  But with RAID-5, the
usable capacity (viewed by the system) of all disk units changes when
units are added or removed from the checksum set.  This adds to the
complexity of attaching a RAID-5 box to an existing system.

      This article describes a hybrid solution that performs better
than RAID-4 and allows growth without changing the usable capacity of
the existing units in the checksum set.

      First, some background information.  A redundant array with
inexpensive disk (RAID) provides protection from data loss in case of
a single unit failure.  Parity in the disk array is used to
reconstruct data.  This improves the availability of the data.

      A redundant array with a single parity unit is called a RAID-4
disk drive (DASD) array.  Data is kept on the N units while parity is
kept on the N+1 unit.  An example with N=3 is shown in Fig. 1.

      The system has addressability to the full capacity of the N
units but cannot address the parity unit.  One advantage of the
RAID-4 architecture is that adding or removing units is relatively
simple because the change does not affect the other N units.  The
parity is simply regenerated for the N units in the set.  The major
disadvantage of RAID-4 is the P unit becomes a performance bottleneck
during write operations.

      A RAID-5 DASD array removes the performance bottleneck by
distributing the parity across all units.  The system has
addressability to all of the units but each has reduced capacity.
The example shown in Fig. 2 has four units, thus the system can
address only 75% (3/4) of each unit.

      The major disadvantage of the RAID-5 architecture is adding or
removing units from the checksum set changes the percentage of user
storage on all the units in the set.  For example, if two units were
added to the configuration shown in Fig. 2, parity...