Browse Prior Art Database

Selecting Mirrored Disk Unit for Read Operations

IP.com Disclosure Number: IPCOM000101788D
Original Publication Date: 1990-Sep-01
Included in the Prior Art Database: 2005-Mar-16
Document File: 2 page(s) / 73K

Publishing Venue

IBM

Related People

Bond, MF: AUTHOR [+5]

Abstract

Disclosed is a means of choosing between two mirrored disk units for read operations in an environment with several layers of I/O control. Taken into account are the relative positions of the disk arms with respect to the I/O request, and the number of I/O commands queued to each disk device of the mirrored pair.

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

Selecting Mirrored Disk Unit for Read Operations

       Disclosed is a means of choosing between two mirrored
disk units for read operations in an environment with several layers
of I/O control.  Taken into account are the relative positions of the
disk arms with respect to the I/O request, and the number of I/O
commands queued to each disk device of the mirrored pair.

      Mirroring disk units is a way of providing system availability,
and it allows smooth recovery from disk failures.  A copy, or mirror,
of each disk unit is maintained by performing each write operation to
two disk units.  While this reduces system performance, an offsetting
gain can be made by judiciously choosing between two disk units of a
mirrored pair for each read operation.  Since the ratio of reads to
writes is approximately 7:3, a small performance gain in read
operations could make up for a larger performance deficit from write
operations.

      There are several layers of software in the path between the
I/O "request" from the operating system to the I/O "command" sent to
the I/O subsystem, to the firmware that is sensitive to physical
properties of the device, such as rotational latency and seek delay.
Selection of the device with the faster expected response is
complicated by these layers of I/O control.  The layer of control
that is described here is somewhere in the middle - in the software
that issues the commands to the I/O subsystem.  The criteria for
selecting which of a pair of units to read from should not interfere
with the scheduling algorithm used by the I/O subsystem.

      The position of each disk arm is recorded by saving the disk
address whenever a read or write command is issued (refer to figure).
When a read request is received, the read is is...