Browse Prior Art Database

Performance Enhancements during Establish Duplex Pair

IP.com Disclosure Number: IPCOM000113763D
Original Publication Date: 1994-Oct-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 138K

Publishing Venue

IBM

Related People

Blount, CL: AUTHOR [+7]

Abstract

One common use of Dual Copy is uninterrupted migration of data when a volume is failing or on the verge of failing. Often the data is on a critical volume and needs to be copied to an alternate volume during a performance sensitive period. To minimize the performance impact, the copy algorithms have been modified to reduce the time required to do the copy, to allow more frequent host access to the device from which the data is being copied and to allow DASD fast write operations during the copy.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 46% of the total text.

Performance Enhancements during Establish Duplex Pair

      One common use of Dual Copy is uninterrupted migration of data
when a volume is failing or on the verge of failing.  Often the data
is on a critical volume and needs to be copied to an alternate volume
during a performance sensitive period.  To minimize the performance
impact, the copy algorithms have been modified to reduce the time
required to do the copy, to allow more frequent host access to the
device from which the data is being copied and to allow DASD fast
write operations during the copy.

      The dual copy function creates and maintains a second logical
image of a DASD volume for enhanced availability.  It is initiated by
the host system through the PSF command or by the SF through a PSF
interface, which initiates the copy of the primary volume to the
secondary volume.  During the copy, while the device is in duplex
pending state, host initiated I/O is accepted and will be queued
until the number of tracks specified by the pace parameter of the PSF
establish duplex command have been copied.

      Under the old algorithms, the physical primary and secondary
devices are reserved for the copy before the specified number of
tracks are staged from the primary and not released until all the
staged tracks have been destaged to the secondary.  During this time,
no physical device selection can occur and no host initiated I/O will
be done.  After the specified number of tracks have been copied, the
host I/O will complete.  If the I/O is a write to a track that had
already been copied, the updated data will be destaged to the
secondary device before the primary device is available to the
channel.  Both the delay before the I/O to the primary device can
complete and the delay while the secondary device is updated for
writes degrade the performance during the copy.

      A new algorithm has been established for copying the tracks
from the primary to the secondary volume.  This new method will allow
more channel access to the device during the copy to improve the
performance of read and write misses.  It will increase the speed of
the copy to decrease the amount of time in duplex pending mode.  It
will also allow DASD fast writes during the copy to improve response
time to the channel for predictable writes and write hits.

      To increase the speed of the copy, the tracks will be staged
from the primary using sequential multi-track prestaging.  The
destage will be done using the multi-track Stage/Destage algorithms.
To allow more channel accesses during the copy, the physical device
as well as the logical device will be available to the channel during
the destage, the logical device will be available to the channel
during the prestage and the physical device will be available more
frequently during the prestage.

      Initially, the first two cylinders in the cylinder bitmap will
be prestaged.  Once a track is staged, it will be freed (set n...