What most of the commenters here are missing is the reality that not every system, function, or business needs the sort of uptime that AWS offers - and that's fine. It's something a lot of newer entrants into the technology field fail to grasp, because they've never had to actually deal with an outage before - or a time when the internet itself was ephemeral and temporary, available only as long as your connection remained active.
The number one thing people poo-pooing these "We saved $XXX by getting off public cloud" posts is that each business has different calculus for its risk tolerances, business needs, and opportunity costs. Once a function reaches some form of stability or homeostasis, then hosting it in the public cloud can become a net liability rather than a net asset. Being able to make those decisions impartially is what separates the genuinely good talent from those who conflate TC with wisdom.
Even when public cloud is the right decision, using managed services increasingly isn't. MongoDB Atlas is a managed service with a corresponding price tag to match. Running it in a VPS like Hetzner may shift some of the maintenance and support tasks onto your team, but let's be real - modern databases are designed to be bulletproof, and huge companies operated just fine with a single database instance on bare metal for decades, even with the odd downtime along the way. We ran a MongoDB CE database at a PriorCo on a single VM in a single datacenter for nearly a decade, and it underpinned a substantial chunk of our operations - operations we could do by hand, if needed, during downtime or outages (that never happened). We eventually moved it to AWS DocumentDB not out of cost-savings or necessity, but because a higher-up demanded we do so.
If anything, the visceral rebuke of anyone daring to move off public cloud feels very reminiscent of my own collegial douchebagginess in the 2000s, loudly mocking Linux stans and proclaiming closed source (Microsoft) would run the planet. Past-me was a douchebag then, and the same applies to the AWS-stans of today.
The number one thing people poo-pooing these "We saved $XXX by getting off public cloud" posts is that each business has different calculus for its risk tolerances, business needs, and opportunity costs. Once a function reaches some form of stability or homeostasis, then hosting it in the public cloud can become a net liability rather than a net asset. Being able to make those decisions impartially is what separates the genuinely good talent from those who conflate TC with wisdom.
Even when public cloud is the right decision, using managed services increasingly isn't. MongoDB Atlas is a managed service with a corresponding price tag to match. Running it in a VPS like Hetzner may shift some of the maintenance and support tasks onto your team, but let's be real - modern databases are designed to be bulletproof, and huge companies operated just fine with a single database instance on bare metal for decades, even with the odd downtime along the way. We ran a MongoDB CE database at a PriorCo on a single VM in a single datacenter for nearly a decade, and it underpinned a substantial chunk of our operations - operations we could do by hand, if needed, during downtime or outages (that never happened). We eventually moved it to AWS DocumentDB not out of cost-savings or necessity, but because a higher-up demanded we do so.
If anything, the visceral rebuke of anyone daring to move off public cloud feels very reminiscent of my own collegial douchebagginess in the 2000s, loudly mocking Linux stans and proclaiming closed source (Microsoft) would run the planet. Past-me was a douchebag then, and the same applies to the AWS-stans of today.