Designating backup and restore responsibilities, Improving performance for backup and restore – Sybase 12.4.2 User Manual
Page 435
CHAPTER 11 Backup and Data Recovery
415
For example, once you have a full backup of your database, in theory you could
perform only incremental backups thereafter. You would not want to do this,
however, because any future recovery would be intolerably slow, and would
require more tape or disk space than doing a full backup periodically.
Remember that other users can have read and write access while you do
backups, but no one else can use the database while you are restoring it. You
might find yourself needing to restore dozens of incremental backups, with
your system unavailable to users throughout the process.
A much better approach is a mix of incremental and full backups.
The greater the volume of your database changes, the more important it is do a
backup, and the smaller the advantage of incremental backups. For example, if
you update your database nightly with changes that affect 10 percent or more
of the data, you may want to do an incremental_since_full backup each night,
and a full backup once a week. On the other hand, if your changes tend to be
few, a full backup once a month with incrementals in between might be fine.
Designating Backup and Restore Responsibilities
Many organizations have an operator whose job is to perform all backup and
recovery operations. Anyone who is responsible for backing up or restoring an
Adaptive Server IQ database must have DBA privileges for the database.
Improving performance for backup and restore
The overall time it takes to complete a backup or recover a database depends
largely on the strategy you choose for mixing full and incremental restores.
Several other factors also affect the speed of backup and restore operations: the
number of archive devices, data verification, the memory available for the
backup, and size of the IQ and Catalog Stores.
Increasing the number of archive devices
BACKUP
and
RESTORE
write your IQ data in parallel to or from all of the
archive devices you specify. The Catalog Store is written serially to the first
device. Faster backups and restores result from greater parallelism. To achieve
greater performance when backing up or restoring a large database, specify
more archive devices.