Rick Cook
Even in enterprises that don't run 24 X 7, the amount of time available for backing up is shrinking as the volume of storage to be backed up grows.
The most extreme solution is to do a hot backup--that is to back up the system while it is still in operation. Some DBMS's, such as Oracle, offer utilities that manage hot backups while maintaining coherency. However hot backups carry penalties. They put an additional load on the network or SAN. As a Sun Microsystems white paper on high-speed backups notes: "Because of the fact that the DBMS must log transactions during the backup process, performance may be degraded--and the larger the database the longer the period over which performance is degraded."
A related solution is to mirror the storage. Essentially this involves running a second storage array that contains a copy of the data. At backup time, the mirrored array is uncoupled from the main array and the backup is taken off the mirror. When the backup is done, the mirror reconnects to the main array and its contents are updated. This isn't as disruptive, especially if the mirror is connected to the backup system by a separate link, but it is expensive because it doubles the amount of storage needed.
The most fundamental strategy for expanding the backup window is to divide and conquer. By using multiple backup devices each serving part of the storage system, backups can be done in parallel. Modern backup software, such as the programs from Legato Systems and Veritas Software Corp. are designed to support multiple streaming backups. While this is cheaper than hot backups or mirroring, it is not without costs. First, of course, are the additional backup devices and the hardware to link them to the storage subsystems. In addition, unless the backup devices have their own links to the storage subsystems they support, the backup speed is limited by the available network bandwidth.
The Sun white paper discussing high-speed backup of large databases is at: http://www.sun.com/storage/white-papers/hspeedbackup.html.
Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.
