It's been five years now. Five years since I've been self-employed, working for clients and on my own projects.
What's interesting to see is that every time people launch projects, it costs them a fortune in infrastructure because they go with Vercel, AWS, or any other PaaS or IaaS.
My vision, from the very beginning, has always been that I needed to reduce these costs. I can't, with every new project, pay Vercel to host it.
I need to be fully capable of managing my own computing power, my storage spaces, and all of that, in order to provide as many services as I want.
I don't want the cost of launching a new service — especially since there will be other costs — to be a barrier.
By default, every new service costs money.
Today, it would be less of a problem, but when I started out, every single euro mattered.
Back then, the company had 500 euros in its account and I was doing Uber Eats deliveries to pay for the servers.
So, for five years, it's been very simple: I use bare metal servers.
I have Ubuntu (10-year support) and Docker containers running on top of it.
I didn't switch to Kubernetes, simply because I had traffic, and the infrastructure as it was could handle that traffic without any issues.
Each service, for example the free services like LFQR and LFNY, is designed with this constraint in mind: what costs money is the infrastructure. That's actually why LFQR and LFNY don't have analytics.
Adding them would require computing power, and I wouldn't be able to provide a free service anymore.
Most importantly, for each hosted service, users couldn't upload their files onto it.
And that's something interesting, because it changes everything.
Now, in 2026, things have changed a lot. As I'm writing these lines, three restaurants have just signed up on ViteUneTable.
The first thing they do is set their theme, their colors, their logo, and the photo of their restaurant.
You might say: that's normal...
Except that now, imagine thousands of restaurants uploading their files. And, since these are restaurant owners who aren't in the tech world, it's not a JPEG or an ultra-optimized WebP — it's a big fat 4-megabyte PNG.
You'll quickly understand that there can be a problem, particularly in terms of costs, pretty fast.
So, I've been turning this problem over in every direction for a month because I know very well that my infrastructure... Within a year, it's over.
There's too much traffic for it to work the way it does today. That's a certainty.
So I've tested a lot of things. I tried a PaaS — it works.
But then you pull out the calculator and you realize that costs explode quickly when new restaurants want to try ViteUneTable and start taking reservations.
Since ViteUneTable offers a free version, that's not an option. I could absorb the costs with my own funds, but that's a long-term bet I won't take.
So, what's my solution going to be? Well, we're going to keep doing what we've been doing for five years, except we're going to do it times ten.
Instead of having two or three servers, we'll have ten.
To orchestrate all of that, we'll place a load balancer in front and Coolify, a software that lets you build a PaaS (it's open source).
It's going to work within a private network — the database servers won't be directly accessible from the Internet, obviously.
The only ones that will have ports 80 and 443 listening on the Internet will be the web servers, and no others.
We're going to keep doing what we've been doing for five years, but we're stepping it up a level.
By the way, the migration has already started. If you're paying attention, you'll notice there's a new subdomain on ViteUneTable called CDN. Yes, it's exactly what you think it is.
What's interesting to see is that every time people launch projects, it costs them a fortune in infrastructure because they go with Vercel, AWS, or any other PaaS or IaaS.
My vision, from the very beginning, has always been that I needed to reduce these costs. I can't, with every new project, pay Vercel to host it.
I need to be fully capable of managing my own computing power, my storage spaces, and all of that, in order to provide as many services as I want.
I don't want the cost of launching a new service — especially since there will be other costs — to be a barrier.
By default, every new service costs money.
Today, it would be less of a problem, but when I started out, every single euro mattered.
Back then, the company had 500 euros in its account and I was doing Uber Eats deliveries to pay for the servers.
So, for five years, it's been very simple: I use bare metal servers.
I have Ubuntu (10-year support) and Docker containers running on top of it.
I didn't switch to Kubernetes, simply because I had traffic, and the infrastructure as it was could handle that traffic without any issues.
Each service, for example the free services like LFQR and LFNY, is designed with this constraint in mind: what costs money is the infrastructure. That's actually why LFQR and LFNY don't have analytics.
Adding them would require computing power, and I wouldn't be able to provide a free service anymore.
Most importantly, for each hosted service, users couldn't upload their files onto it.
And that's something interesting, because it changes everything.
Now, in 2026, things have changed a lot. As I'm writing these lines, three restaurants have just signed up on ViteUneTable.
The first thing they do is set their theme, their colors, their logo, and the photo of their restaurant.
You might say: that's normal...
Except that now, imagine thousands of restaurants uploading their files. And, since these are restaurant owners who aren't in the tech world, it's not a JPEG or an ultra-optimized WebP — it's a big fat 4-megabyte PNG.
You'll quickly understand that there can be a problem, particularly in terms of costs, pretty fast.
So, I've been turning this problem over in every direction for a month because I know very well that my infrastructure... Within a year, it's over.
There's too much traffic for it to work the way it does today. That's a certainty.
So I've tested a lot of things. I tried a PaaS — it works.
But then you pull out the calculator and you realize that costs explode quickly when new restaurants want to try ViteUneTable and start taking reservations.
Since ViteUneTable offers a free version, that's not an option. I could absorb the costs with my own funds, but that's a long-term bet I won't take.
So, what's my solution going to be? Well, we're going to keep doing what we've been doing for five years, except we're going to do it times ten.
Instead of having two or three servers, we'll have ten.
To orchestrate all of that, we'll place a load balancer in front and Coolify, a software that lets you build a PaaS (it's open source).
It's going to work within a private network — the database servers won't be directly accessible from the Internet, obviously.
The only ones that will have ports 80 and 443 listening on the Internet will be the web servers, and no others.
We're going to keep doing what we've been doing for five years, but we're stepping it up a level.
By the way, the migration has already started. If you're paying attention, you'll notice there's a new subdomain on ViteUneTable called CDN. Yes, it's exactly what you think it is.