Thousands of popular websites went dark Tuesday morning. An hour later, they were back online. But the incident, linked to cloud provider Fastly, illustrates how fragile internet infrastructure is.
The White House. The New York Times. Tech giant Amazon.
Those were just three of numerous institutions whose websites went black Tuesday.
Early in the morning, visitors started seeing error messages when they tried to access the websites of the British government, the social media platform Reddit, or media outlets like the BBC, CNN and the Financial Times.
An hour later, the pages came back online. But the damage was already done.
The outage could have cost companies hundreds of millions in lost revenue, after consumers were unable to access websites and online shops, analysts estimate.
The incident was linked to a network failure at cloud computing company Fastly. The US-based firm helps websites speed up traffic by hosting some of their content.
In an email, a Fastly spokesperson confirmed that the company had found a problem in its system that "triggered disruptions." She did not provide details about what exactly caused the incident but said that the company had fixed the bug.
The outage illustrates that, as the technology that underpins the internet's infrastructure is increasingly run by a small number of companies, this leaves the web vulnerable to disruptions.
"A few companies now host a lot of stuff … and the list has got smaller over the years," said Doug Madory, director of internet analysis at network analytics firm Kentik, which analyzed the incident.
"If one of these big guys — it may be rare — but if they do have an outage, this takes a lot of things down,” he added.
"A few companies now host a lot of stuff … and the list has got smaller over the years," says one expert.
Widely unnoticed, little-known intermediaries like Fastly have become pillars of internet infrastructure.
Based in San Francisco, the company operates a global network of servers and data centers. In those centers, Fastly stores copies of content from websites it cooperates with. When visitors in a certain region access the pages, they're shown the replicated data instead of the original. This speeds up the loading pace of the websites.
During Tuesday's outage, however, traffic running through Fastly's servers seems to have been interrupted. An analysis by analytics firm Kentik suggests it plummeted by 75 percent.
Internet outages are no new phenomena. In recent years, software glitches in digital infrastructure have repeatedly brought down the internet.
In 2017, human error was responsible for an outage at Amazon's cloud service AWS, which shut down websites for hours.
Last December, a glitch brought down services of US tech giant Google. And in January, users of the business messaging app Slack were unable to send messages or make calls.
What made Tuesday's outage worrisome, however, is its scale.
It affected thousands of websites around the globe, including France's Le Monde newspaper and Sweden's social insurance agency, according to Reuters.
And the incident will likely not be the last internet outage, experts believe.
"As much as we in the internet industry try to completely eliminate these types of errors that are self-inflicted, you can't completely eliminate them," said Kentik's Doug Madory, "There will be outages in the future again."