Views:

Summary



This article describes how image backups work.

Resolution



Image backup is not available for all operating systems, since it relies on the availability of a native file system snapshot capability.

In an image backup, DPX performs a snapshot-based, high-performance backup. This mode maximizes backup efficiency especially for backup clients that store large numbers of files. An image backup dramatically reduces the number of disk I/Os and CPU cycles required for a backup by creating an image of data blocks only and not free space. An image restore supports restoration to the file level, even though the backup of the underlying file system data blocks was done as a stream rather than by backing up files one by one.

An image backup consists of the following actions. If any of these actions fail, the image backup fails.

  1. The Nibbler module initiates a file system snapshot using the operating system or a third party snapshot driver.
  2. The Nibbler module executes a separate module to build the table of contents from the raw data in the snapshot image.
  3. The Nibbler module performs a raw physical backup on the snapshot image device. Currently on Win32 NTFS, only the allocated blocks are processed. Otherwise, all blocks are processed.
  4. After a successful backup phase, file history is sent to the DPX master server using the table of contents data from step 2. Although the actual data was backed up like a raw backup of the disk, the file history information saved to the catalog provides the ability to restore the data as if it were backed up from the individual files.
  5. The device snapshot is destroyed.

Only nodes that support image backup can be selected when defining an image backup job.

As with other backup job definitions, an image backup can be run as a base, incremental, or differential backup. The sequence of actions described above applies to the base backup type only. When the backup is an incremental or differential type backup, the file system will be read file by file just like a regular file system backup, with no snapshot or raw physical backup activity. This is required because for these backup types, the files will be backed up based on their timestamps, indicating that they have changed since the last backup of the relevant type. These will be no different in operation from a backup defined to use the file system.