Self-Managed VPS vs Managed AWS

Most VPS-versus-AWS comparisons put one hosting bill next to another. On that basis AWS loses almost every time, and the comparison is worthless. A $20 server is cheaper than a $200 one. Nobody needed a tool to work that out.

The number that actually decides it is your own admin time. Patching, certificates, backups, the evening a disk filled up. That work is real and it costs real money, but it never appears on an invoice, so it gets left out of the decision. This tool puts it in.

So the comparison runs on three things: what you pay for hosting, how many hours a month you put into it, and what an hour of your time is worth. Change the third one and the answer can flip completely. That is not a flaw in the tool. That is the actual shape of the decision.

What each thing you enter does

• Monthly hosting cost. Your side of the ledger, unchanged. Include backups, panel licences, anything billed monthly.
• Admin hours a month. Multiplied by your rate, this becomes the largest line in most results. Routine work only — incidents are asked separately so they are not double-counted.
• Your hourly rate. The multiplier on everything above. It starts empty and stays empty until you fill it in.
• Incidents a year, and hours each. Priced as your time, then carried identically on both sides. Moving platform does not fix what causes incidents, so this line cancels out of the recommendation entirely. It is shown because it is real money, not because it argues for anything.
• What is yours to do. The 8 responsibilities. Each one carries the share that this AWS shape actually removes, and two of them remove nothing at all.
• Backups and how it is run. These change no numbers. They decide which problems get named, and those are true whichever platform you land on.
• The AWS sizing. Roughly what you would need over there. Kept short on purpose — the assumptions filling the gap are all printed with the result.

What you stop being responsible for, and what you do not

This is the part worth reading twice, because it is where these comparisons usually go wrong. Managed AWS is bought for the work it takes off you, not for the compute. But it does not take off as much as people assume.

RDS genuinely owns your database engine, its patching and its minor-version upgrades. ACM issues and renews certificates on the load balancer, for free, and does not forget. Multi-AZ genuinely fails over without you. Those are real, and they are most of the case for moving.

However: EC2 is a server you rent. It does not patch itself. It does not deploy your application. If those two are where your hours go, moving to AWS moves them across intact, and you get a larger bill for the privilege. The tool scores both at zero for that reason, and it will happily tell you the move saves you nothing.

There is a floor under the AWS side, and it is higher than people expect. Size the smallest thing anyone would sensibly run — one small instance, no managed database — and the AWS bill still lands around forty dollars a month before a minute of your time is counted. More than half of that is the load balancer.

So if you are paying five or ten dollars for a VPS and spending twenty minutes a month on it, AWS does not win that comparison. Push your hourly rate up to two hundred dollars and the best it manages is a draw. The tool will tell you so, and you should stop reading and go and do something more useful.

Roughly speaking, below about an hour of admin a month a small VPS wins outright. Between one and two hours it is genuinely level. Above that the answer starts depending on what you tick, and the break-even line tells you exactly where it turns. The honest version of this tool is one that loses the sale often, and this one does.

For what it is worth: this site runs on a self-managed VPS, on DirectAdmin, patched by me. Every tool on it, including this one, is served from that box. I did not move it to AWS because for what it does, the numbers on this page say not to.

Comparing the two invoices and stopping there

This is the big one. An invoice comparison is a comparison of the cheapest part of running a server. Price the hours first, then look at the invoices, and the argument often reverses. Sometimes it reverses back — but at least it was the real question.

Assuming "managed" means nobody patches anything

It means nobody patches the database. Your operating system, your web server and your application are all still yours on EC2. If you were hoping to stop doing Linux administration, this particular shape of AWS is not how you do that.

Forgetting NAT Gateway and egress

A private database subnet needs a NAT Gateway, which bills by the hour and again per gigabyte through it. Traffic leaving AWS is billed on the way out too, so the same gigabyte can be charged twice. Most VPS plans bundle bandwidth, so this is genuinely new money rather than a like-for-like line. It is switched on here by default rather than quietly off.

Expecting a migration to fix the incidents

It will not. Whatever is filling the disk, leaking memory or failing to come back after a restart will do the same thing on a different provider's hardware. That is why this tool prices incidents on both sides identically. Fix the cause, then decide where to run it.

Counting an untested backup as a backup

If nobody has restored one, you do not know that you have backups. You know that you have files. Restoring the smallest one into a scratch machine takes an afternoon and either confirms it or finds the problem on a Tuesday instead of during an outage. Do that before you spend anything on a migration.

Thinking one instance is safer than one server

It is not. One EC2 instance in one availability zone is one failure domain, exactly like one VPS. Redundancy on AWS is something you design and pay for. Availability is not a property you get by changing supplier.