It only took a technical blunder at a little-known Internet company to shut down a host of the world’s best-known websites Tuesday morning.
A software glitch on a content delivery network run by San Francisco-based Fastly blocked online access to the New York Times, CNN, and the BBC, as well as the popular forum Reddit, videogame streaming service Twitch, and a number of other websites.
The disruption is the latest in a series of major Internet disruptions with national and global impact. Cyber criminals shut down a major US fuel pipeline in May, causing serious gasoline shortages, while a similar attack last week briefly closed down meatpacking plants that produce much of the nation’s beef, pork, and chicken.
The Fastly outage wasn’t a deliberate attack, and it lasted for less than two hours, but it shows how vulnerable we all are to single points of failure on the Internet. The company first detected a problem at 5:58 a.m. Eastern time, according to its network status website, and installed a fix for the problem at 7:58 a.m. “We identified a service configuration that triggered disruptions across our POPs (points of presence) globally and have disabled that configuration,” said an e-mail message from Fastly at 9:44 a.m. “Our global network is coming back online.”
So why did it happen? A company like the Times, which distributes large amounts of data worldwide, doesn’t want to run its own massive central data center to handle all the traffic. Instead, it contracts with Fastly, which operates a global network of servers in major population centers. Digital copies of the Times are stored throughout the Fastly network and are quickly delivered to local users.
“Instead of building one giant gas station in the middle of the state, we have lots of smaller gas stations located in your neighborhood,” said Andy Champagne, a senior vice president at Fastly’s biggest competitor, Cambridge-based Akamai Technologies.
Champagne said that companies can try to protect themselves by signing up with two content delivery network services, so they can remain online if one goes down. But he said that this solution is costly and complicated. It also requires a computer system that shares content between both networks, a single point of failure that could still cause a total network shutdown if something breaks.
“Any time you add redundancy, you also add complexity,” said Champagne. “Technology is always going to have problems.”