Incremental snapshot mechanism

Last Updated: Nov 30, 2017

Snapshots on Alibaba Cloud are taken using an incremental method. In this method, two snapshots are compared and only the data that has changed is copied, as shown in the following image:
1

In the preceding figure, Snapshot 1, Snapshot 2, and Snapshot 3 are the first, second, and third snapshots of a disk. The file system checks the disk data by blocks. When a snapshot is created, only the blocks with changed data will be copied to the snapshot. In this example:

  • In Snapshot 1, all data on the disk is copied since it is the first disk snapshot.
  • Snapshot 2 only copies the changed data blocks B1 and C1. Data blocks, A and D, are referenced from Snapshot 1.
  • Snapshot 3 copies the changed data block B2 but references data blocks, A, D, from Snapshot 1, and references C1 from Snapshot 2.
  • When you roll back the disk to Snapshot 3, blocks A, B2, C1, and D are copied to the disk, to replicate Snapshot 3.
  • When you delete Snapshot 2, block B1 will be deleted, but C1 will remain because blocks that are referenced by other snapshots cannot be deleted. When you roll back to Snapshot 3, block C1 will be recovered.

When the disk needs to be restored to the status at the time of Snapshot 3, you can perform snapshot rollback to copy data blocks A, B2, C1, and D to the disk.

If Snapshot 2 is deleted, data block B1 in the snapshot will be deleted but data block C1 will not be deleted. In this way, when the disk is restored to the status at the time of Snapshot 3, data block C1 can also be restored.

Snapshot creation time varies depending on actual volume.

Snapshots are stored on the Object Storage Service (OSS), but they are invisible to users and will not be computed in the OSS space occupied by the users’ buckets. Snapshot operations can only be performed through the ECS console or APIs.

Thank you! We've received your feedback.