The way your backup strategy is designed is based on the requirements for your data recovery. These requirements can simply be formulated as "How much information are you willing to lose, and what is the maximum time available for recovery?" These questions boil down to financial costs. The perfect solution is one of no loss of information and recovery in zero seconds. Such a solution requires significant investments in hardware to accomplish this level of data protection. The result is choosing the optimal solution for your organization, meaning that your recovery strategy balances the costs of losing information and the investments you need to make to recover as quickly and with a little data loss as possible. In fact, it all has to do with risk management and contingency planning.
A well-established backup and recovery strategy is essential to your organization. This is also the case for the backup of your Windows 2000 installation and registry, applications, Active Directory database, and user data. However, you should head for the data vault to retrieve the tape only if all other options fail.
The architecture of the complete IT infrastructure should be designed and implemented with a high level of redundancy offering fail-over capabilities, especially if you are running 24 x 7 applications like many of today's Web applications. Within this kind of infrastructure, you should consider a number of options. The first is redundancy of the database server using solutions such as Microsoft Cluster Services, so your database servers fail over in the event of system problems and your applications can continue to operate. Within your server, all parts such as network cards and power supplies should be in a redundant configuration, which guarantees continuity of operation.
A second option in maintaining SQL Server availability is the use of SQL Server 2000 capabilities such as log shipping and replication. These techniques safeguard your system from unnecessary loss of data or availability by having warm or replicated standby servers.
The third option involves dispersing your data across multiple drives. Never configure backup files stored to disk, transaction logs, system databases, and user-defined database files on the same physical disk. Besides offering performance improvements, a dispersed configuration increases the restore capabilities and recovery time over a single physical disk configuration.
Remember that relying on media backups is the same as accepting loss of data as media backups will never contain the most recent data in online transaction processing systems. Since the loss of data is unacceptable in most cases, maintaining your servers for high availability should be your goal. Having to go to your backups to recover your server should be your last option.
There are three types of problems in which the need for recovery is necessary: permanent loss of one or more servers due to a disaster (natural or manmade); failure of storage devices—for example, a disk crash or power failure; or invalid data modifications (human error)—for example, a content update batch run twice, making a large number of records invalid.
Was this article helpful?