Also because it's cheap.

That's a bonus!

Same for me with OVH and Scaleway.

Way cheaper than AWS and friends, and they just work.

I'm using OVH and the notion of it "just working" is all relative. It's tolerable, but certainly a bit buggy, and with far less services than aws and co. It is also cheap, but given the limitations I doubt they can increase the prices much...

Missing object storage kinda makes it a joke. Hetzner cloud is certainly useful for some things, but object storage is something I just assume any cloud would have in 2024.

AWS launched it before EC2 even it's that valuable.

Out of curiosity are you operating at a scale where "MinIO on a Hetzner node" isn't a viable replacement? I totally believe that's possible, just curious about the use case.

Are you serious? Self-managed MinIO on a bunch of drives vs fully managed operation-less S3. Comparing apples to oranges. And the license, AGPLv3 is a non-starter for virtually any business serious about their intellectual property.

> AGPLv3 is a non-starter for virtually any business serious about their intellectual property.

What do you mean? Can you describe a scenario where this MinIO license matters?

I'm confusing things myself. One's okay as long as one doesn't modify the code.

If you thought "I had to publish any code changes I make to my object storage server" was bad, wait until you find out that you're not even allowed to make code changes to S3 servers!

Obviously the fully-managed nature of S3 is very valuable compared to MinIO but the licensing issues seem neither here or there.

I’d argue that the smaller you are, the more sense S3 makes, since its costs scales down to zero and offer a high degree of reliability from the first file stored.

Rolling your own storage comes with at least some operational expenses, if you’re counting your data/egress in gigabytes that’s likely not worth it compared to just using S3 or similar services.

S3 is so expensive at scale that it only makes sense if you're tiny or if you have extreme durability requirements and your working set is tiny.

The durability can have value at scale, because ensuring it at scale with MinIO can be a lot work. But I often recommend to clients that they deploy a Hetzner setup as a cache, often with storage all the way up covering 100% of their data depending on access partner, because the AWS egress fees are so insanely extortionate that you never really want to read data from AWS at scale other than as an emergency option.

I just use R2. Who says that if I use one company, they have to provide every single service under the sun to be useful?

Linode is quite cheap and comes with Object storage so you don't have to maintain a MinIO

Yeah, that - and also managed Kubernetes. Hetzner could seize a lot of potentially lucrative opportunities, but for some reason they choose not to and pretty much stagnate in their offerings. I have been asking the support about both K8s and object storage since close to five years now, but no.

I also used to use Hetzner and I got the feeling they were capitalised to the extent their backers understood the DC/asset model and under-capitalised to software which was too intangible to invest in.

K8s probably means more s/w than hardware, more bodies (for them) more helpdesk and more documentation, more process. More up-front cost. And, they seem to be making money without it. Maybe thats their position?

Their position is that k8s is the wrong layer of abstraction. They might be right, for their business model. Or it might be a case of intentional blindness.

It is cheap, good it is not.

You can complain about AWS and GCP support as much as you want, but given their sheer volume it is an accomplishment how rarely you actually need it. The situation when you need it is of course a different thing and I understand that satisfaction there greatly varies. Their support might be hit or miss, but at least their processes are streamlined and tried and tested enough to have you rarely need it.

With Hetzner on the other hand, if you use them seriously you will run into a situation where you need support sooner than later and my experience with that always has been abysmal.

Do you have some examples?

I do. My favourite is the one where a scraper hosted their version of Stackoverflow on Hetzner servers. For some reason they only copied the textual content but replaced all user profile pictures with random profiles pictures from elsewhere - mostly harmless ones. I had the misfortune that mine was replaced by a pornographic one, which appeared as the first result in Google when someone searched my name.

Hetzner did nothing to help rectify the situation. Thanks for nothing Hetzner.

What I find great about the Hetzner Cloud is that you don't need a dictionary to understand which product does a specific job. With AWS, I found it much harder to take the first steps because of the names being used.

I find AWS quite OK (Stockholm syndrome?), but Cloudflare dashboard confuses me every time. Their navigation is just not logical to me.

I like to think UpCloud (also EU-based) is in the same boat, servers are servers and databases are databases. No weird lingo or product names, at least when you get to actually using it all in the Hub (hub.upcloud.com)