Why is the internet broken? How one [company?*] can take down the whole web – UK Times

It started in a computer in northern Virginia, at around 3am local time. But just a few moments later it was everywhere.

When an Amazon data centre stopped working properly on Monday morning, so did much of the internet. For hours, users were unable to access many of the world’s most popular apps, platforms and services: everything from Snapchat to the UK’s tax authority were down.

The outage reflects the complex nature of today’s web: deeply intertwined with people’s lives, at once resilient and fragile. For critics, the dramatic, sudden and widespread nature of the incident shows that we rely too heavily on a small number of companies to power the internet, though the fact that it is so rare and was fixed so quickly also shows exactly how those companies were given that power in the first place.

Amazon Web Services describes itself as “providing industry-leading cloud capabilities and expertise”. “Amazon Web Services is the world’s most comprehensive and broadly adopted cloud, enabling customers to build anything they can imagine,” its website reads. “We offer the greatest choice of innovative cloud capabilities and expertise, on the most extensive global infrastructure with industry-leading security, reliability, and performance.”

The jargon and puffery might be a little much, but it’s also a reflection of the importance and complexity of Amazon’s business. In more straightforward terms, Amazon Web Services offers the infrastructure that powers the web: computers, data centres and other technologies that are hired out to companies so that they can show us the Snapchat messages and games that we ask for.

It has been a very successful business. As could be seen in the widespread nature of the outage, many companies rely on AWS: it is the biggest company of its kind in the world, competing with Microsoft Azure and Google Cloud, and the money it makes now makes up the majority of Amazon’s profits.

That growth has come in large part because it makes life so much easier for companies. When they are building an app or a website, they don’t need to commission data centres filled with computers at the same time – they can instead just hire them as they need them from Amazon, which in turn is able to be more efficient as a result.

It also means that it is largely reliable. Amazon’s terms require it to try and ensure that its services are up at least 99.99 per cent of the time, and it achieves that aim.

But the tiny sliver of time in which it breaks – and when that in turn breaks the internet – can be disastrous. That means that outages of this kind often lead to more philosophical questions about whether the internet’s infrastructure should be more broadly spread out.

The problem at Amazon Web Services appears to be related to the Domain Name System, or DNS, usually referred to as the “phone book of the internet”. When a person types an easily-remembered web address into their browser – www.the-independent.com, for example – DNS is used to translate that into the actual address where the required website can be found.

The way that works is a reflection of the communitarian, perhaps somewhat utopian vision of the internet that guided the early years of its development. The system is spread across the web and across the world, allowing for the interlinked nature of the internet to work.

But it is exactly that sense of a distributed web that has been somewhat lost in recent years – and which means that outages like those seen at AWS today are rare but dramatic when they happen. The technologies that underpin the internet have become concentrated in a relatively small number of companies, allowing for increased efficiency and economies of scale but also balancing a large part of the internet on a relatively small surface.

Even companies that are big enough to operate their own infrastructure have run into problems with centralisation. In recent years, for instance, Meta has come to bring together the technology that powers its app – Facebook, Instagram, WhatsApp and more – so that they are much the same on the back end. But that has meant that those rare outages at the company bring down all of its products, at the same time.

And the centralisation problem isn’t limited to web infrastructure. Last year, when a bad update from Crowdstrike was pushed to Windows computers, it took down a whole host of things that aren’t directly related at all: hospitals, airlines and physical stores. But they had all been using the same cybersecurity protections, and so all broke at the same time.