Amazon S3 Lets Customers Ship Big Data

IDG News Service — Amazon's S3 cloud storage service has a new option, called AWS Import/Export, for quickly uploading large amounts of information to its data centers. It uses a well-developed, multimodal content delivery network that can transmit terabytes of data faster than a T-3 leased line.

The fact that this network is based on jets, trucks and messengers with walkie-talkies doesn't make it any less useful to enterprises, many of which have been using overnight shipping services for backups for several years, according to 451 Group analyst Henry Baltazar. Just make sure the data's encrypted in case it falls off the back of a truck or otherwise gets lost, he said.

AWS (Amazon Web Services) described the new service in a recent blog posting. AWS Import/Export from Amazon Web Services lets customers send virtually unlimited amounts of data to Amazon when they want to start using S3 for the first time, back up their content offsite, or streamline the Direct Data Interchange process with their partners. All customers will have to do is copy their data to a device, such as an external hard drive, create a manifest file with authentication information and a digital signature, e-mail loading instructions and ship the device. AWS lays out guidelines for the storage devices on an information page at its Web site.

When it arrives, the device will go to an AWS Import/Export station and the data will be loaded onto the customer's S3 data bucket, generally the next business day. Customers will pay US$80 per device handled and $2.49 per hour for the labor involved in loading the data, plus the standard charges for storing that data on S3. The service is available now in beta testing, for importing only, but will be expanded to include exporting in the coming months, Amazon said.

With many enterprise Internet connections, Import/Export will often be faster than online uploads or downloads, according to Amazon. For example, on a T-1 leased line (1.5Mb per second), with 80 percent of that line devoted to the transfer, it would take 82 days to send 1TB of data, Amazon said. As a general rule, S3 customers with only a T-1 should think about using Import/Export for sending 100GB or more of data, the company said.

Even a faster T-3 leased line (just under 45Mb per second) would take three days to send 1TB, so shipping would be a good option for anything above 2TB, Amazon said. A Gigabit Ethernet Internet connection could send 1TB in less than a day, Amazon said. But even if an enterprise is using a metro Ethernet link like that, it's unlikely to have that amount of capacity all the way to Amazon, 451's Baltazar pointed out.

"If you really have to have that data up there fast, it does make sense," Baltazar said. The method isn't new: For example, when banks set up new branches and want to have large amounts of information available on site, they typically ship drives because they don't have days to wait for a transfer. Online backup and disaster-recovery vendors also offer this approach. It's developed in just the past few years as the growth in data, driven by multimedia, has outpaced the acceleration of Internet connections, he said.

What's new is that Amazon, a cloud storage provider that offers more than just backup, is using the technique. The core business model for AWS is providing storage on S3 for applications that run on Amazon's EC2 cloud computing infrastructure.