The Magnificent Cloud: Too Big To Fail... *yet it does

AWS outages highlight the risks of single-provider cloud reliance. Explore why multi-cloud resilience is vital for digital infrastructure.

The cloud, a ubiquitous term for the countless unseen and powerful servers that deliver our digital existence, promises omnipresence and perfect reliability, yet remains profoundly fragile.

Yesterday’s widespread AWS outage, caused by a DNS failure in its critical US-EAST-1 region, demonstrated this vulnerability once again. When a single component in one of the world’s largest public clouds fails, the effects are felt across our digital lives.

This wasn’t an isolated incident, and it certainly won’t be the last. The history of this ‘Magnificent Cloud’ is littered with similar events, from the widespread 2021 AWS outage to non-AWS events like the 2024 CrowdStrike software update that crippled systems globally. Each failure is a stark reminder that while the cloud is vast and somewhat ethereal, its critical points of failure are few and concentrated in the hands of a small number of providers.

The Concentration of Digital Power

Our digital lives, from basic websites to cutting edge web services, including AI generative models and autonomous agents are increasingly hosted on the platforms of a few major players. Looking at the “Big Three” - Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP) - they collectively run over 60% of the global cloud infrastructure market. AWS holds the largest share, hosting in access of 70 million websites therefore powering a huge swathe of the internet including Netflix, Spotify and Airbnb.

This dependency is both good and bad for the providers and the organisations that deploy on them. While it cements their market dominance, it also makes their failures existential threats. AWS is the financial powerhouse at Amazon, often accounting for half or more of the company’s total operating income despite representing a smaller portion of overall revenue. Its high profit margin has made it a pillar of investor confidence. Yet, the relative calm in the stock market following major outages suggests investors have become complacent, viewing these disruptions as momentary hiccups rather than systemic risks. Why?

The Perils of Single Provider Reliance

If the risk is so evident, why don’t more organisations implement redundancy across multiple cloud providers? The reasons are myriad, but a blend of technical complexity, financial incentives and operational complacency:

• Complexity and Skill Gaps: Managing different APIs, security protocols, and operational dashboards across AWS, Azure, and GCP (and others) requires a highly specialised and expensive skill set. Interoperability between different cloud platforms remains a major technical challenge.
• Cost Management: While the desire should be to spread risk, multi-cloud is obviously more expensive. Companies can receive substantial volume discounts from a single provider. Furthermore, managing unpredictable costs across multiple vendors adds financial overhead.
• Operational Ease: It’s simply easier for billing and management to deal with one vendor, despite the inherent risk of vendor lock-in.

A reluctance to invest in proper multi-cloud infrastructure means that when an outage occurs, companies, government services and users suffer. While users may laugh that they can’t get their Snapchat fix, the reality can be a massive loss of productivity and seen on the bottom line from missed sales. When organisations are increasingly dependent on sophisticated AI tools and cloud-based services lose access to their digital ecosystems the impact is significant.

A Call for Resilient Architecture

What if yesterday’s event hadn’t been an internal DNS mistake (human error) but an act by a rogue actor? We’re seeing corporate attacks by cyber criminals growing exponentially - in the last month we’ve seen the ongoing attack at Jaguar Land Rover in the UK, and Asahi in Japan.

Think about how often you use a digital service in your average day. The widespread loss of access to say websites, financial services and critical government services highlights how exposed our collective lives are to a single point of failure.

Over the years we’ve thought about this a lot as a team at Anomify. We recognise that the only responsible path forward is to factor in redundancy from the ground up. Sure, it costs us more, and is far more complex, but ironically it is what helps us sleep better at night. We even patented our approach to this, which we call ZPF (Zero Point of Failure). Our solution addresses this problem by enabling true failover delivery across multiple cloud providers. We know we’re not alone in recognising that the industry must pivot from simply mitigating failures and waiting it out, to architecting for real uptime.

We’re always surprised (or not) when we see the high profile names which go down when these incidents happen. While absolute constant uptime remains a myth, a failure to deploy with high resilience and multi-cloud failover in mind is a real problem.

The cloud is magnificent, but the digital economy has grown too big to fail on its infrastructure. It’s time to stop gambling on single-source reliability and begin building better; truly resilient, distributed web services we require for the future.

This time the AWS outage was largely resolved within a few hours, but what about next time?