Skip to content
PuglieseWeb Engineering HandbookSystemDox Pages

Volume Gateway

Volume Gateway presents block storage volumes over iSCSI and backs them with Amazon S3. It is the gateway type for applications that need a block device rather than a file share — databases, backup software, anything that formats its own file system. See the Storage Gateway overview for how it compares with the file and tape types.

Two modes

Section titled “Two modes”

Cached mode. The primary copy of the data lives in S3; frequently accessed blocks are cached on local disk. On-premises storage requirements drop to the working set, which is what makes this the mode used during a migration and in cost-driven hybrid architectures.

Stored mode. The primary copy lives on local disk and is replicated asynchronously to S3. All reads are local, so latency is predictable, and S3 holds an off-site copy for backup and disaster recovery rather than serving reads.

The distinction is which copy is authoritative, and it decides how much local capacity is needed and what happens when the link to AWS goes down.

Snapshots

Section titled “Snapshots”

Volumes support point-in-time snapshots, taken on demand or on a schedule. Snapshots are stored as Amazon EBS snapshots, which is the property that makes Volume Gateway a disaster-recovery tool rather than just a backup target: a snapshot can be restored directly as an EBS volume and attached to an EC2 instance, so the workload can be brought up in AWS without a separate conversion step.

flowchart LR subgraph OnPrem["On-premises"] APP[Application server] VGW[Volume Gateway] DISK[(Local disks<br/>cache or primary)] end subgraph AWS["AWS Cloud"] S3[(Amazon S3)] SNAP[EBS snapshots] EBS[EBS volume on EC2] end APP -->|iSCSI| VGW VGW <--> DISK VGW -.->|Async upload| S3 S3 --> SNAP SNAP -->|Restore| EBS style VGW fill:#FF9900,stroke:#232F3E,color:white style S3 fill:#3F8624,stroke:#232F3E,color:white

Common uses

Section titled “Common uses”
  • Migration. Start in stored mode so nothing changes for the application, let the data synchronise to S3, then switch to cached mode and reclaim the local array.
  • Disaster recovery. Keep snapshots current, and recover into EC2 from EBS snapshots rather than rebuilding from tape.
  • Backup and restore for applications whose backup software expects block devices.