Amazon Glacier: Low-Cost Cloud Archiving, But Not If Data Needs Occasional Access

Amazon Web Services on Tuesday unveiled a new public cloud, long-term archiving solution that allows customers who are not concerned about high-speed access to data to do their archiving in the cloud for as little as 1 cent per GB of data per month.

The move could spell trouble for vendors of hardware and software used for archiving data, including tape vendors, given the cost of the new service and Amazon's promised data durability.

AWS said the aptly-named Amazon Glacier is not for users in a hurry. A single file or multiple files can be uploaded into Amazon Glacier as an archive, with access requiring the initiation of a job. A job typically takes 3.5 hours to 4.5 hours, AWS said.

[Related: Cloud Storage For All: How To Build Your Own Practice ]

It could be considered the opposite of AWS's new Amazon Elastic Compute Cloud instance called High I/O, which provides an SSD front-end to the AWS service for applications requiring high-speed, low-latency access to data in the cloud.

Amazon Glacier is slated to be a highly reliable service. AWS claims average annual durability of 99.999999999 percent, or 11 nines of availability, which is the same durability as that of its Amazon S3 standard storage. Amazon estimated that a durability of 99.999999999 percent means that a customer might expect to lose a single data object once every 10,000,000 years.

Amazon Glacier secures the data by transferring it over Secure Sockets Layer (SSL) and encrypts the data using AES256 encryption.

The cost of the service is as low as 1 cent per GB of data per month, with no up-front commitments or operational expenses required, which means users only pay for what capacity they use. However, the company will charge an additional 3 cents per GB for items that are deleted within 90 days of being moved to the service. In addition, there is a 1 cent per GB for any amount of data over 5 percent of the amount of data stored with Amazon Glacier, pro-rated on a daily basis.

Amazon Glacier can be used with other AWS offerings, including AWS Import/Export to allow movements of large amounts of data into the service via portable storage devices. Amazon S3 is expected to have an option for moving data between Amazon S3 and Amazon Glacier based on users' data lifecycle policies, the company said.

NEXT: Better Archiving Choices For E-discovery, Searches?

Tape is currently the most commonly used media for archiving. LTO-5 tape costs as low as 3 cents per GB with 2:1 compression. When used for archiving purposes, the tape, once the data has been written, can be taken offline and stored without incurring future power costs.

However, there are additional costs related to the long-term storage and transport of tapes, as well as questions about how long a tape can be expected to reliably store data for long-term archiving.

AWS said that tape or other media, when used for archiving, requires the upfront purchase of the array or library, as well as on-going operational costs including power, facilities, staff and maintenance. Those costs are not incurred with Amazon Glacier, the company said.

The value of Amazon Glacier depends on the use case, said Greg Knieriemen, an enterprise technology podcaster and vice president of marketing at Chi Corp., a Cleveland, Ohio-based solution provider.

At 1 cent per GB per month, it is cheaper to buy and use a hard drive for many archiving cases, Knieriemen said.

"For long-term retention, Amazon Glacier's biggest benefit will be for customers who never need to access their data again," he said.

For most of Chi's customers, archived data will be accessed once stored, Knieriemen said.

"If you retain data for compliance, your compliance policies and how you access the data dictates how it is used," he said. "If you do a lot of e-discovery and searches, I'm not sure Amazon Glacier is a good option. Many of our customers use archiving for infrequently-accessed data, but they still access it."

In addition, archiving data is more than just dumping the data somewhere, Knieriemen said. "How do you manage it," he said. "Search and e-discovery requires active access to the data."

PUBLISHED AUG. 21, 2012