As you’ll already know, Amazon Web Services (AWS) suffered a major web outage this week, with internet performance tracker Ookla reporting over 16 million user outage reports globally across more than 60 countries.

This week also saw Arelion expand its Baltic network infrastructure, bypassing Russia in a region it dubbed “geopolitically sensitive.” Monday meanwhile saw Vodafone Ukraine announce a submarine cable system spanning the Black Sea, also bypassing Russia.

What’s the connection? I’ll give you a clue: it begins with the letter D.

We’re talking dependency.

Humpty Dumpty had a great fall

But first, let’s start with another D word, and that’s DNS (Domain Name System).

Much has been made of the enforced AWS brain drain, which has seen the cloud giant cut its engineering staff in recent years in favor of an AI-augmented workforce. One dev, Justin Garrison, posted at the end of 2023 his experience at the hyperscaler, suggesting service teams were either laid off or forced to quit.

Garrison also made this canny prediction:

“I suspect there’ll be a major AWS outage in 2024. No amount of multi-region redundancy will protect you.”

He was off by about a year, and didn't pin down the outage to a source such as DNS. But the argument remains solid. In this case, Humpty Dumpty AWS had a great fall, and there weren’t enough king’s horses and king’s men to quickly fix it.

In the end, it took 15 hours to solve the failover. But would more horses and men have prevented the fall? Did AI fail to help in this instance, or will AI do a better job than fauna in the future?

Debatable. The problem at the end of the day isn’t intelligence, but complexity. With increased scales comes an increased fragility, and in this case a single failure in DNS led the house of cards to fall.

The real problem, arguably, is dependency.

Backbone sovereignty is the fix

To understand my point, it’s time for a D-led term, and that’s data sovereignty.

The European data residency question has recently led Cisco, SAP, SUSE, Microsoft, OVHcloud, Exoscale, and more to invest in EU-compliant solutions, as the EU Data Act kicked into gear last month to weaken vendor lock-in for Europe's cloud customers.

Now, data sovereignty isn’t the magic bullet against cloud outages. That’s because they are often the result of control-plane, shared-service, or DNS issues. These aren’t fixed by simply hosting one’s data locally.

But a dependency on hyperscalers is perhaps as wise as depending on a major aggressor in your neighborhood to provide part of your energy and network needs. That’s why Arelion and Vodafone have made their investments in encouraging diversity in the Baltic and Black Sea regions.

While a hyperscaler is neither a nation state nor a malign influence on the world, its royal status can inadvertently make a weapon of sorts through failure. And those depending on it may pay the price.

As Tech Policy Press put it, “the outage once again demonstrates how concentration in the computing industry … can crash major parts of the internet.

“These disruptions are not just technical issues; they are democratic failures,” wrote authors Corinne Cath and Don Le.

Tech EU concurred, dubbing resilience “as much a geopolitical issue as a technical one,” while Computing UK said the AWS failure highlighted the “urgent need for multi-region resilience and sovereign cloud strategies.”

Some initiatives, such as Gaia-X, promote a federated and secure framework for sovereign data exchange, aiming to reduce dependence on the American cloud giants.

But again we’re in software territory when we should be looking at the physical, tangible efforts being made in Russia’s backyard as a source of inspiration.

True backbone sovereignty requires diversifying where your services are run, owning your control plane, bolstering network paths, and running fire drill-like practices in the event of a failover.

Backbone sovereignty would also ensure legal obligations of vendors, and perhaps even reclaiming control of the network spine. If that isn’t possible, then build it back better.

But for this to work would require a lot of people to put their back into it. Some may even need to grow a proverbial spine.

It would require a No Kings moment of sorts.

Until then, sit back and get used to outages. The AWS failover is only the beginning.