Docs

Elastic Compute Service

Snapshots

How snapshots work

A snapshot is a backup of a volume taken at a specific point in time. This topic explains how snapshots work, including incremental snapshots, rollback, and deletion.

Incremental snapshots

Snapshots use an incremental approach to save storage space and costs. The first snapshot you create for a volume is a full snapshot. Subsequent snapshots are incremental, meaning they only capture data that has changed since the last one.

As shown in the diagram below, snapshots 1, 2, and 3 represent three different points in time of a volume.

Snapshot 1 is the first snapshot of the volume. It backs up all the volume's data blocks.

Snapshot 2 is the second snapshot of the volume. It only backs up the changed blocks (A' and D'). Blocks B and C reference data from snapshot 1.

Snapshot 3 is the third snapshot of the volume. It backs up only the changed block (B''). Blocks A' and D' reference data from snapshot 2, and block C references data from snapshot 1.

Rolling back a volume

When you roll back a volume using a specific snapshot, EBS automatically merges it with the previous ones. If there are identical data blocks, the system retains the data from the most recent snapshot. During the rollback process, EBS writes the merged data to the volume.

As shown in the diagram below, snapshots 1, 2, and 3 represent three different points in time of a volume. When you roll back using snapshot 3, the system restores the volume to the state of the data captured in that snapshot, which consists of blocks A', B'', C, and D'.

Deleting snapshots

You can delete some of your snapshots to free up space if you no longer need to keep them or if you have more than the limit allows.

As shown in the diagram below, snapshots 1, 2, and 3 represent three different points in time of a volume. Deleting snapshot 1 will delete blocks A and D, which contain new data, while blocks B and C, which are still referenced, will be retained.

Last updated: 2026.05.15 15:10:44