Tuesday, December 02, 2008

Amazon Aims at Content Delivery

Startups may be drawn to the company's content-delivery network.

By Erica Naone --- from technologyreview.com

Credit: Technology Review

Amazon has become a major player in cloud computing in recent years. Many Web startups have come to rely on its pay-as-you-go hosting and computing services rather than investing in costly and complex infrastructure of their own. The newest offering from Amazon Web Services, called Cloudfront, may provide insight into Amazon's long-term business model. This new product offers companies that are already hooked on Amazon storage and processing the ability to distribute their content and thus make themselves more stable and reliable.

Cloudfront is a distributed content-delivery network. It improves the performance of a website by strategically placing copies of that site's content on servers around the world. A user of such a site who is located in Europe, for example, might see an improvement in performance by loading that data from a European server, rather than from the original server hosted in the United States.

The giants of content delivery--companies like Akamai and Limelight--need a lot of infrastructure, not to mention clever algorithms, to provide services to massive Internet companies such as Google and Microsoft, and this makes it difficult for new entrants to challenge them. Amazon has servers dotted around the world and says only that it has "invested significantly" in building capacity for Cloudfront.

However, rather than take on the giants of content delivery directly, the new service is a fairly simple add-on to the company's existing storage solution, Amazon S3 (Simple Storage Service). "[Customers] asked us for a service that complemented Amazon S3's high-durability storage with even higher performance delivery, by storing and delivering popular content from edge locations close to where their customers make requests," says a company spokesperson. "That is why the service was built."

Cloudfront reflects the fact that it is basically an extension of an existing solution in that it offers a fairly basic content-delivery service. The infrastructure can be used to distribute large files but not live streaming video, and it is not as large or complex as the content-delivery network offered by Akamai, for example. However, Amazon says that other features are on the way, again as an outgrowth of existing uses of Amazon infrastructure. "We intend to add streaming in later versions of the service," the spokesperson says. "Many Amazon Web Services customers already use Amazon EC2 [Elastic Compute Cloud] to stream content live or on demand."

The customers that Amazon has chosen to test Cloudfront give some clue to the target audience. Woot, an e-commerce site that offers a different deal every day, is using the service to cope with spikes in traffic that occur right when a new deal is posted. "We're not [photo-sharing site] Flickr," says Woot's retail IT director Luke Duff. "We're just doing our sale images for products. As far as performance, we only really care about the current day's items." Since Woot already used other features of Amazon Web Services, he says, it just made sense to try Cloudfront, especially since the cost was "quite a bit less" than other services that the company investigated. And because of the site's volatile traffic patterns, pay-as-you-go works well for Woot, Duff adds.

Amazon isn't the only company hoping to fill this niche in content-delivery networks with a pay-as-you-go business model. Voxel, based in New York, has offered, for about a year, pay-as-you-go content delivery, including support for streaming video. Voxel CEO Raj Dutt says that he expects pay-as-you-go content delivery to appeal particularly to customers with unpredictable traffic patterns.

James Staten, a principal analyst at Forrester Research, believes that Cloudfront is likely to attract startups that are new to using content-delivery networks. But this might not be a bad thing for more-established players. "If I was Akamai, I wouldn't be too worried," Staten says. "Because [Cloudfront] is a limited implementation, it becomes a nice on-ramp for customers, when they mature, to move up to Akamai." Growing companies may eventually move on, he says, because "the value of a content-delivery network is directly related to how many locations it provides you." More-established content-delivery companies are currently much better than Amazon at providing sophisticated access to a wide variety of locations, Staten says.

But even though Amazon Cloudfront may be less sophisticated, Staten says that Amazon's popular S3 and EC2 products are "absolutely" going to draw companies into trying Cloudfront and any other add-ons that Amazon develops in the future. "If you're drinking the Kool-Aid and the experience is good," Staten says, "there's no point in not trying the additional services that they make available to you."