Three services that all "move data" but do completely different jobs. The whole topic is knowing which job each one is for.
They get confused because all three touch "data going to AWS." Separate them by what job and one-time vs ongoing:
| Service | The job it does | One-time or ongoing? | How |
|---|---|---|---|
| Storage Gateway | On-prem apps keep working but use AWS as their storage backend. | Online (local cache) | |
| AWS Backup | Scheduled, centralized backups + retention across AWS services. | Online (policy-driven) | |
| Snow Family | Move a huge one-off pile of data into AWS when the network can't. | Offline (physical shipping) |
Storage Gateway = ongoing hybrid access · AWS Backup = scheduled backups · Snow Family = one-time bulk migration by truck/box.
A gateway appliance (VM/hardware) sits on-prem, presents the storage protocol your apps already speak, caches hot data locally for low latency, and syncs everything to AWS in the background. Your apps don't know the data went to the cloud.
| Type | On-prem speaks… | Stored in AWS as |
|---|---|---|
| S3 File Gateway | NFS / SMB (files) | Objects in S3 |
| FSx File Gateway | SMB (Windows shares) | Amazon FSx for Windows |
| Tape Gateway | Backup software (virtual tape / VTL) | S3 + Glacier (replaces physical tape) |
| Volume Gateway | iSCSI (block volumes) | S3 + EBS snapshots |
Volume Gateway modes: Cached = primary data in AWS, only hot data cached on-prem · Stored = entire dataset on-prem, async backups to AWS as EBS snapshots.
"Keep using the on-premises app but store/back up the data in AWS" → Storage Gateway. Families = File / Tape / Volume.
One policy-driven place to automate and centralize backups across many AWS services, instead of scripting snapshots per service. Built for retention rules, compliance, and disaster recovery.
"Centrally manage / automate backups across multiple AWS services" or "meet a backup retention / compliance policy" → AWS Backup (+ Vault Lock for immutable, ransomware-proof backups).
AWS Backup is for scheduled protection & retention, not bulk one-time migration (that's Snow) and not live hybrid access (that's Storage Gateway).
Physical, rugged, encrypted devices AWS ships to you. You load your data, ship it back, and AWS imports it to S3. For when uploading over the internet would take weeks or months. Sizes: cone < ball < mobile.
| Device | Capacity | Form factor | Best for |
|---|---|---|---|
| Snowcone | ~8 TB (HDD) / 14 TB (SSD) | Smallest, rugged, fits in a backpack | Small/edge transfers, tight or harsh spaces; light edge compute. |
| Snowball Edge | ~80 TB (use several for PB) | Suitcase-sized | TB–PB transfers. Storage-optimized for data; compute-optimized (vCPU/GPU) for edge processing. |
| Snowmobile | up to 100 PB | 45-ft shipping container on a truck | Exabyte-scale whole-data-center migrations. |
If transferring it over your internet connection would take weeks+, use Snow. Size by volume: TB / portable → Snowcone · ~80 TB–PB → Snowball · exabytes / data-center move → Snowmobile.
| When the question says… | Answer |
|---|---|
| "keep using on-prem app, store/back up in AWS", hybrid, local cache | |
| "replace physical tape backups" | |
| "centrally automate backups across AWS services" | |
| "immutable / ransomware-proof backups for compliance" | |
| "move 50 TB, slow internet" | |
| "small/edge transfer, portable rugged device" | |
| "migrate an entire data center, exabytes" | |
| "run compute on the device in a disconnected location" |