Backup and Restore: Classification of Scenarios
|
|
|
In this article
Backup approach
First, there are two main approaches to performing backup.
- File-level backup includes only certain files from your environment. It may or may not include some high-level file system data such as file permissions or alternate file streams.
- Application-aware backup is a special case of file-level backups that hides underlying file structure from administrator and operates in terms of a specific application instead. For example, a "database-aware" backup could include database file and transaction logs files as a single unit. This eases administration and helps ensure consistent data recovery. It may also provide additional features like consistency check of backups or truncating transaction logs after successful backup iteration.
- System State backup is another special case of file-level backup. It provides a predefined set of operating-system-specific files that are crucial for OS health. See the next sections on how you can benefit from this type of backup.
- Volume-level backup (also called "imaging") backs up a data volume (also called a "logical disk" or "partition") as a whole. Often that is done on per-sector mode that it includes all file and underlying file system information. However, some imaging applications and formats capture data at the file level instead, and may not include file system information.
Some modern backup applications track file or volume changes at the block level and are able to minimize bandwidth as well as backup media usage storing only changed blocks of protected files or volumes. Though that does not turn file-level backup to volume-level or vice versa.
The choice between file-level backup and volume-level backup should drive your choice of backup software. Most backup software implements one of these types of backup but not both. Because it is usually impractical to perform both types of backup for the same data, it is not uncommon to combine different backup types for different kinds of data in the same IT environment.
Restore scenarios
Usually you don't care much on what you back up and how. What you are really worrying about is what you can restore and how. There are generally two types of restore operations you may need.
- Item-level restore is something you want when a user deletes a file on error or you need to recover a previous version of a single SQL Server database per request of an auditor.
- Disaster recovery is something you need when entire server, its storage unit or the whole data center was crashed or burned in fire. Please note that the term “Disaster recovery” sometimes includes having a backup data center but we're not going to discuss this here. All we care about at this point is what we need to restore, not where.
Obviously the backup solution (or combination of solutions) you chose should support both restore scenarios.
Restore procedures depending on backup type
If you have file-level backup and need to perform an item-level restore you're lucky. That's the easiest scenario usually. You just restore the data you need and to original or alternate location. Depending on type of application that works with this file you may also need to perform additional steps. E.g. if you restored an older copy of production database to an alternate location you probably would need to attach it manually before it becomes available to the auditors.
If you have file-level backup and find yourself in a disaster you probably should reinstall operating system (OS) and applications and only after that restore data. Performing unattended installation, scripting and/or having System State backup can greatly enhance these procedures and help eliminate human errors. Post-recovery manual steps can also be required or scripted.
If you have volume-level backup and need to perform an item-level restore everything depends on the capabilities of backup software you use. Most of modern backup applications provide features called “Item-level recovery (ILR)” or the like. Typically it mounts (or somewhat extracts) the image to the backup server and allows you to select individual files or other items (such as database units) for restore. If your backup application does not provide this feature you need to restore the whole image to alternate location (unused disk volume or LUN) and copy necessary files manually.
If you have volume-level backup and need to recover from a disaster you need the feature usually referenced as “Bare-Metal Recovery (BMR)”. This allows you to recover entire volume to new (or recovered) piece of hardware and turn it into exact copy of failed (or lost) server(s). Technically that usually means booting into special recovery environment (using “Live CD” or network boot) and applying the image. Alternatively you could recover only data volume if the OS is not affected by the failure or you prefer to reinstall it. This does not need BMR feature and usually can be performed by backup software installed.
Related titles
- Before learning about special features of certain backup and restore applications we recommend you read about [[articles:Backup and Restore: Special Considerations|Special Considerations for Backup and Restore]].