The hour-long Internet outage, which began at around 11am BST, was revealed by Fastly to be due to issues with its global content delivery network (CDN), which underpins many major sites.
Outages were reported in locations such as London, Texas and New Zealand, while Berlin remained unaffected.
In a statement, the company said: “We identified a service configuration that triggered disruption across our POPs (points of presence) globally and have disabled that configuration.
“Our global network is coming back online.”
As well as streaming sites such as Hulu and Twitch, the incident also affected the presentation of emojis on Twitter, with the alt text replacing images.
Visitors to these sites received error messages, including “Error 503 service unavailable” and “connection failure”.
Fastly operates an edge cloud approach, which is designed to accelerate loading times for websites, as well as protect against denial-of-service (DOS) attacks.
Content delivery networks: Six criteria for selecting a suitable CDN
Following the outage, Gaz Jones, co-founder and technical director of digital agency, Think3, said: “Fastly CDN had major problems affecting Stack Overflow, Spotify, Stripe, Gov.uk and GitHub among others.
“This is what happens when half of the internet relies on Goliaths like Amazon, Google and Fastly for all of its servers and web services. The entire internet has become dangerously geared on just a few players.”
Similar worldwide outages have impacted the operations of Amazon Web Services (AWS) and Google in the past.
Reliance on CDNs
Toby Stephenson, CTO at Neuways, believes today’s Internet outage “highlights the reliance of many of the world’s biggest websites on CDNs such as Fastly”.
Stephenson continued: “As there are so few of these CDN services, these outages can occur from time-to-time. By using these CDNs to push content to readers, these websites are usually fast and responsive, but on this occasion they have been left with egg on their collective faces.
“The technical backends of these big websites are probably fine, but it is the frontends that can’t be accessed, and content cannot be pushed as the network is down.”
