AWS Snow Family
Key concepts
Snowcone for edge
Snowball Edge Storage/Compute
Snowmobile for exabytes
Data migration use cases
Edge computing capabilities
Overview
The AWS Snow Family is a set of physical devices that AWS ships to your location so you can move large amounts of data into and out of AWS when the network is too slow, too expensive, or unavailable. Instead of pushing terabytes or petabytes over the internet, you copy data onto a rugged appliance and ship it back to AWS, which loads the data into Amazon S3 (Simple Storage Service, the object storage service). The family includes AWS Snowcone (the smallest, portable edge device) and AWS Snowball Edge (Storage Optimized and Compute Optimized variants). AWS Snowmobile, a shipping container pulled by a truck for exabyte-scale migrations, was retired in 2024 and is no longer available for order, but it still appears in older exam-prep material.
This topic carries a low weight on the SAA-C03 exam, but it appears in scenario questions about data ingestion and migration. You should recognize when offline transfer beats online transfer, match the right device to a data volume, and know that Snow devices can run compute (Amazon EC2 instances and AWS Lambda functions) at the edge in disconnected or rugged environments such as factories, ships, and remote sites. Note that AWS now offers Snow devices only to existing customers and points new customers to AWS DataSync and AWS Data Transfer Terminal, though the offline-transfer concepts remain testable.
Offline Transfer for Big Data and Bad Networks
Use the Snow Family when transferring large datasets over the network would take too long or cost too much. Snowcone handles up to about 14 TB at the edge, the current Snowball Edge Storage Optimized device holds about 210 TB, and the retired Snowmobile handled up to 100 PB for exabyte-scale data center migrations. Snowcone and Snowball Edge can also run EC2 and Lambda compute locally.
Match the device to the data volume: Snowcone for a few TB and portable edge use, Snowball Edge Storage Optimized for about 210 TB per device (and Compute Optimized for edge processing or an optional GPU), Snowmobile for up to 100 PB in legacy scenarios. A common signal is when an online transfer would take weeks or months: choose Snow. Data is encrypted with AWS KMS and loaded into S3 on return.
Key Concepts
Snowcone for the Edge
The Smallest, Portable Device
AWS Snowcone is the smallest member of the family, a rugged device small enough to fit in a backpack and light enough to carry by hand. It comes in an HDD model (about 8 TB of usable storage) and an SSD model (about 14 TB of usable storage). Snowcone runs at the edge with limited compute (it can host EC2 instances and AWS IoT Greengrass for local processing) and is designed for harsh, space-constrained, low-bandwidth locations such as vehicles, drones, and field sites. Data moves to AWS either by shipping the device back or by using AWS DataSync (an online data transfer service) over an available network.
Snowball Edge Storage and Compute
Two Snowball Edge Variants
AWS Snowball Edge comes in two configurations. The Storage Optimized variant maximizes capacity (about 210 TB of usable NVMe storage in the current generation) and is the workhorse for bulk data migration. The Compute Optimized variant provides 104 vCPUs and 416 GB of memory with about 28 TB of NVMe SSD, and offers an optional GPU, so it can run heavier edge workloads such as machine learning inference and video analytics. Both variants can run EC2 instances and Lambda functions locally, which lets you process data on the device before it ever reaches the cloud.
Snowball Edge Variants
| Dimension | Storage Optimized | Compute Optimized |
|---|---|---|
| Primary purpose | Bulk data migration | Edge compute and processing |
| Usable storage | About 210 TB NVMe | About 28 TB NVMe SSD |
| Compute | Moderate (EC2 and Lambda) | 104 vCPUs and 416 GB memory |
| GPU option | No | Yes (optional) |
| Typical use | Large one-time transfers | ML inference, video analytics at the edge |
Snowmobile for Exabytes
A Data Center on a Truck
AWS Snowmobile was a 45-foot ruggedized shipping container hauled by a semi-trailer truck, capable of moving up to 100 PB (petabytes) of data per Snowmobile. It was built for the largest migrations, such as relocating an entire on-premises data center or a video archive to AWS. AWS personnel drove it to your site, connected it to your network for high-speed transfer, and returned it to an AWS Region for loading into S3. It included physical security measures such as GPS tracking, 24/7 video surveillance, and an optional security escort. AWS retired Snowmobile in 2024, so treat it as a legacy exam concept: when a scenario describes exabyte-scale moves today, the practical answer is multiple Snowball Edge devices or a high-bandwidth online path.
Data Migration Use Cases
When Offline Beats Online
The core decision is transfer time. A useful rule of thumb: moving 100 TB over a dedicated 100 Mbps connection takes roughly 100 days, while a Snowball Edge ships in about a week. Choose Snow devices for one-time bulk migrations, data center decommissioning, large media and genomics archives, and any site where bandwidth is limited, metered, or unreliable. For ongoing or incremental online transfer over an existing network, AWS DataSync is the better tool, and AWS Direct Connect (a dedicated network link) suits continuous high-volume transfer.
Choosing a Transfer Method by Volume
| Data Volume | Recommended Approach |
|---|---|
| Up to a few TB at a remote edge site | Snowcone |
| Tens to hundreds of TB per shipment | Snowball Edge Storage Optimized (about 210 TB each) |
| Petabytes to exabytes (data center scale) | Multiple Snowball Edge devices (Snowmobile in legacy questions) |
| Ongoing online sync over a usable network | AWS DataSync |
| Continuous high-volume online transfer | AWS Direct Connect |
Edge Computing Capabilities
Run Compute Where the Data Is
Snowcone and Snowball Edge are more than storage boxes. They can run Amazon EC2 instances and AWS Lambda functions locally, so you can filter, transform, compress, or analyze data on site before shipping or uploading it. Compute Optimized Snowball Edge devices add an optional GPU for tasks like real-time image recognition. This matters for disconnected or intermittently connected environments (ships at sea, oil rigs, military field operations) where sending raw data to the cloud first is impractical. The devices integrate with AWS IoT Greengrass for local IoT processing.
Snow Family at a Glance
| Device | Capacity | Compute | Best For |
|---|---|---|---|
| Snowcone | About 8 TB HDD or 14 TB SSD | Limited (EC2, Greengrass) | Portable edge, harsh sites |
| Snowball Edge Storage Optimized | About 210 TB usable | Moderate (EC2, Lambda) | Bulk migration |
| Snowball Edge Compute Optimized | About 28 TB NVMe, 104 vCPUs | High (optional GPU) | Edge ML and analytics |
| Snowmobile (retired 2024) | Up to 100 PB | Transport only | Exabyte data center moves |
Best Practices
1. Estimate transfer time before choosing online vs offline
└── If an online copy would take weeks or months, ship a Snow device
2. Match the device to the data volume
├── Snowcone: a few TB and portable edge use
├── Snowball Edge Storage Optimized: about 210 TB per shipment
└── Petabyte data center scale: parallel Snowball Edge devices (Snowmobile in legacy questions)
3. Use edge compute to shrink what you transfer
└── Run EC2 or Lambda on the device to filter and compress first
4. Order multiple Snowball Edge devices in parallel for very large jobs
└── Several devices can move petabytes when a single appliance is too small
5. Rely on built-in encryption and tracking
├── Data is encrypted with AWS KMS keys
└── E Ink shipping labels and the device auto-update destinationsCommon Pitfalls
Pitfall 1: Using Snow When the Network Is Sufficient
Mistake: Ordering a Snowball Edge for a few gigabytes or for data that the existing network could upload in hours.
Why it fails: Shipping and handling add days of delay and cost for volumes that online transfer would finish quickly.
Correct Approach: Use AWS DataSync or a direct S3 upload for small or network-feasible volumes; reserve Snow devices for large datasets or poor connectivity.
Pitfall 2: Picking Snowmobile for Tens of Terabytes
Mistake: In a legacy exam scenario, choosing Snowmobile to migrate a few tens of terabytes of data.
Why it fails: Snowmobile was engineered for petabyte-to-exabyte data center relocations, so it is oversized and over-provisioned for small jobs.
Correct Approach: Use one or more Snowball Edge Storage Optimized devices for tens to hundreds of terabytes, and reserve the exabyte-scale Snowmobile answer for 10 PB and larger migrations in older questions.
Pitfall 3: Overlooking Edge Compute on Snow Devices
Mistake: Treating Snowcone and Snowball Edge as storage-only and shipping raw, unprocessed data.
Why it fails: Moving raw data wastes capacity and ignores the ability to process on site, which is critical in disconnected environments.
Correct Approach: Run EC2 instances or Lambda functions on the device to filter, transform, or run inference locally, and choose Compute Optimized when you need a GPU.
Test Your Knowledge
A research vessel collects 20 TB of sensor data per voyage in an area with no reliable internet connection, and the team needs to run image-recognition inference on the data while still at sea. Which option fits best?
An enterprise is shutting down a data center and must move 8 PB of archived data to Amazon S3 as a one-time migration. Online transfer would take many months. Which Snow Family answer does the classic exam scenario expect?
A field team needs a rugged, backpack-sized device to gather about 5 TB of data at remote drone-survey sites with limited power and bandwidth. Which device should the architect recommend?
Related Services
Quick Reference
Device Limits and Capacities
Snow Family Quick Facts
| Device | Max Capacity | Compute | Encryption |
|---|---|---|---|
| Snowcone HDD | About 8 TB usable | Limited EC2 and Greengrass | AWS KMS |
| Snowcone SSD | About 14 TB usable | Limited EC2 and Greengrass | AWS KMS |
| Snowball Edge Storage Optimized | About 210 TB usable | Moderate EC2 and Lambda | AWS KMS |
| Snowball Edge Compute Optimized | About 28 TB NVMe, 104 vCPUs | High, optional GPU | AWS KMS |
| Snowmobile (retired 2024) | Up to 100 PB | Transport only | AWS KMS |
Common CLI Commands
# Create a Snowball import job (data moves from device into S3)
aws snowball create-job \
--job-type IMPORT \
--resources '{"S3Resources":[{"BucketArn":"arn:aws:s3:::my-migration-bucket"}]}' \
--address-id ADID-1234567890 \
--kms-key-arn arn:aws:kms:us-east-1:123456789012:key/abcd-1234 \
--role-arn arn:aws:iam::123456789012:role/snowball-import-role \
--snowball-type EDGE_S \
--shipping-option SECOND_DAY
# Check the status of a job
aws snowball describe-job --job-id JID-1234567890
# List address records used for shipping devices
aws snowball describe-addresses
# Get the unlock code and manifest to access the device
aws snowball get-job-unlock-code --job-id JID-1234567890