Amazon, Google, IBM, Microsoft: how their data-migration appliances stack up

Need to move petabytes of data into the cloud? There’s an old-school option that works faster than network transfers: Load a physical appliance with data, then ship the appliance to a cloud provider.

Large-scale data migrations can incur high network costs and unrealistically long transfer times. Even fast networks can take months and years to move terabytes and petabytes of data from an on-premises data center to a public cloud provider.

To make data transport easier, cloud providers have created physical storage devices that are built to travel. An enterprise can set up an appliance in the data center, load it with content, and then ship the appliance to the cloud storage company. A FedEx truck driving down the highway doesn’t face the bandwidth constraints of conventional network links.

Amazon was the first of the cloud giants to offer a shippable storage device for migrating petabyte-scale data sets into the cloud. Its AWS Snowball appliance is a ruggedized device that resembles a PC tower. Google entered the market with its Transfer Appliance, a rackable storage server that started shipping this summer. IBM offers the Mass Data Migration device, and Microsoft has the Azure Data Box and newly previewed Azure Data Box Disk.

Enterprises can use these appliances to transfer large quantities of data – such as analytics data, video libraries and image repositories – to cloud storage. An enterprise might be migrating applications to the cloud, archiving inactive data, decommissioning a data center or collecting data from a research partner or vendor for analysis.

Data replication is another potential use case: An enterprise could transfer a copy of its data to the cloud for redundancy and recovery, for example, or for machine learning and analysis. And some of the appliances work for transferring data out of a cloud environment, too: If there’s a disaster, for example, an enterprise could use an appliance to retrieve data stored in the cloud.

These rugged devices also are targeted at enterprises that operate in remote areas with little or no access to high-speed Internet.

“One should never underestimate the bandwidth of a semi truck filled with disks hurtling down the highway,” AWS sums up. “These methods are ideal for moving large archives, data lakes or in situations where bandwidth and data volumes are simply unrealistic.”

The data-transfer appliances from Amazon, Google, IBM and Microsoft differ in size and capacity; here are the specifics.

AWS Snowball

The ruggedized Snowball device includes a 10GBaseT network connection, and all data that’s transferred to the appliance is secured with 256-bit encryption. It includes an E Ink shipping label for tracking. Enterprises can opt to use a single Snowball device to transport multiple terabytes of data, or use multiple devices in parallel to transfer petabytes of data (plus there’s the AWS Snowmobile truck-size option for exabyte-scale jobs).

AWS

AWS charges a service fee for each data transfer job ($200 for a 50TB Snowball device and $250 for an 80TB Snowball device). The deal includes 10 days of onsite usage with each extra onsite day costing $15. Shipping charges depend on carrier rates, location and priority. It’s free to transfer data into Amazon S3; data transfers out of S3 range from 3 cents to 5 cents per gigabit, depending on location.

Google Transfer Appliance

Google recommends its Transfer Appliance for companies that want to move more than 20 terabytes of data or for data that would take more than a week to upload to Google Cloud Platform. It’s designed to be mounted in a standard 19-inch data center rack, and the ata is encrypted as it’s captured. Once the device is shipped back to Google and the data is uploaded to the cloud, the customer can log into the console to decrypt the data and select storage for it.

Google

Transfer Appliance comes in two sizes: a 2U device with 100TB capacity and a 4U design with 480TB capacity. The 100TB model has a usage fee of $300 and comes with 10 days of onsite usage; each extra day costs $30. The 480TB model has a usage fee of $1,800 and comes with 25 days of onsite usage; each extra day costs $90. Two-day shipping via Fedex is approximately $500 for the 100TB model and $900 for the 480TB model.

IBM Mass Data Migration

IBM’s Mass Data Migration device is designed to be tamper-evident, waterproof and shockproof for protection during device handling and transport. Each 55-pound device on wheels offers up to 120 terabytes of usable capacity; enterprises can use multiple devices to accommodate larger workloads.

IBM

Inline data compression, AES 256-bit encryption and RAID-6 configuration are standard.

In the U.S., IBM charges a flat rate of $395 per device, which includes a $295 device fee, $100 roundtrip overnight delivery, and 10 business days of use. Each extra day of on-site use costs $30. Mass Data Migration is designed to offload data into IBM Cloud Object Storage. It does not support exporting data out of the IBM cloud at this time.

Microsoft Azure Data Box

Microsoft launched a preview program for the Azure Data Box in late 2017, and in July it took the wraps off the Azure Data Box Disk.

The ruggedized, tamper-resistant Data Box weighs in at 45 pounds, can store approximately 100 terabytes, and supports 256-bit encryption.

Microsoft

The newly previewed Azure Data Box Disk is an SSD-disk-based option that’s designed for lower capacity jobs, such as such as moving data from remote or branch offices into Azure.With Data Box Disk, enterprises can order up to five 8-terabyte disks, for a total capacity of 40 terabytes per order. Data is protected using 128-bit AES encryption.

Microsoft charges discounted preview pricing for Data Box, which is available in the U.S., European Union and United Kingdom. Roundtrip device shipping is $95, and the import service fee is $125. Each extra day of on-site use costs $7.50.

The Data Box Disk preview is available in the U.S. and European Union. The Data Box Disk preview is currently free. Microsoft plans to release pricing information later this year.