Amazon Web Services (AWS) this week launched its Glacier cloud storage service as a low-cost service for archived...
data that's rarely accessed but needs to be retained for long periods. It's very different than the established Amazon Simple Storage Service (Amazon S3), which is for data that needs to be accessed in real-time.
Amazon Glacier is an alternative to tape archives, and is designed for data that can tolerate a three- to five-hour retrieval time. Amazon will charge 1 cent per gigabyte (GB) per month to store data on Glacier versus S3's starting price of 12.5 cents per GB per month.
"Glacier is meant for storage that's infrequently accessed; and when it is accessed, the customer and application can tolerate latency," said Alyssa Henry, vice president of AWS Storage Services. "S3 is for applications that are accessed frequently and need to be retrieved in milliseconds."
More on Amazon cloud storage services
Tutorial: Amazon Web Services security
Amazon Web Services program explained
Digital media archives, financial and health care records, raw genomics sequence data, long-term database backups and regulatory compliance information are the most likely data types to end up on Glacier. Henry said Amazon is working to integrate the two cloud services so data stored in Amazon S3 can be migrated to Glacier as it ages via a policy. "We expect to have that in the coming months," she said.
For data retrieval, the first 5% of data downloaded per month is free with a charge of 1 cent per gigabyte after that. "Peak usage is factored in," Henry said. "If 10 requests are done in one hour, it will cost more; but if requests are spread over time [it's cheaper] because you're not creating a lot of activity. The system is designed for cold storage, so we expect data to be rarely retrieved. It will be a very rare case that a customer needs to download more than 5% of data per month."
Similar to Amazon S3, data stored in the Amazon Glacier cloud storage service resides in multiple data centers in multiple regions. Amazon is promising that files stored via Glacier have an annual durability of 99.999999999%, which means companies can expect to lose one of 100 billion stored objects each year.
"Theoretically, data loss is expected to be extremely low," said Andrew Reichman, senior analyst at Cambridge, Mass.-based Forrester Research Inc. "People are talking about how you would manage 100-year archiving and not lose data. That's a concern. In 100 years, technology changes tremendously. By taking this data on, Amazon is committing to moving that data to new technology and they're signing on to deal with that issue."
Amazon Glacier can be set up from the AWS Management Console where customers can upload any amount of data they need. The service is available now.