Netflix Outage, and the Limits of the Internet

Netflix customers hoping to stream a movie with their family this Christmas may have had to turn to watching the Yule Log channel instead, courtesy of a widespread Netflix outage that the company is blaming on its cloud computing service provider Amazon Web Services. 

Heavy demand on Christmas caused the breakdown at Amazon’s data center in northern Virginia. As the New York Times explains, the problem was likely related to a failure of Amazon’s “elastic load balancing” service that spreads high traffic loads across many different servers. Netflix had also prepared measures to prevent such outages, such as buying excess server capacity to prepare for spikes in activity, and storing data in different “zones” to minimize the impact of any given failure. These didn’t work on Monday night. 

An emerging concept called content-centric networking (see “A Rewired Internet Would Speed Up Content Delivery”) might have helped Netflix and Amazon in exactly this situation. Today, Internet traffic is shuffled around in small packets of data between networked IP addreeses that connect a home computer to a specified location, like an Amazon server. Instead, content-centric networking would work by requesting a particular piece of content by name and retreiving it from nearby routers that have handled that same data before. The delivery of videos, especially videos that many people are watching (say, It’s a Wonderful Life), could get faster and more efficient.  

Creating this new vision for the Internet would cost a lot, and likely won’t be economical soon, as my colleague Tom Simonite reported. But the explosion of video-watching and other high-bandwidth Web and mobile uses could one day demand it.