Dedicated servers on a host like Hetzer and OVH surely beats any virtualization based cloud offering on price and performance. The tradeoff is availability. It's a great choice for entities that are optimizing on cost, but not a great choice if your business cannot tolerate downtime.
A good example is a the big lichess outage from last year [1]. Lichess is a non-profit, and also must serve a huge user base. Given their financials, they have to go the cheap dedicated server route (they host on OVH). They publish an Excel sheet somewhere with every resources they use to run the services and last year, I had fun calculating how much it would cost them if they were using an hyperscaler cloud offering instead. I don't remember exactly but it was 5 or 6x the price they currently pay OVH.
The downside, is that when you have an outage, your stuff is tied to physical servers and they can't easily be migrated, when cloud provider on the opposite can easily move around your workload. In the case of Lichess outage, it was some network device they had no control of that went bad, and lichess was down until OVH could fix it, that is many hours.
So, yes you get a great deal, but for a lot of businesses, uptime is more important than cost optimization and the physicality of dedicated servers is actually a serious liability.
> It's a great choice for entities that are optimizing on cost, but not a great choice if your business cannot tolerate downtime.
Even hosting double of everything when you're doing dedicated servers will let you have cheaper monthly bills, compared to the same performance/$ you could get with AWS or whatever.
But Hetzner does seem a bit worse than other providers in that they have random failures in their own infrastructure, so you do need to take care if you wanna avoid downtime. I'm guessing that's how they can keep the prices so low.
> is that when you have an outage, your stuff is tied to physical servers and they can't easily be migrated
I think that's a problem in your design/architecture, if you don't have backups that live outside the actual servers you wanna migrate away from, or at least replicate the data to some network drive you can easily attach to a new instance in an instant.
Hetzner only has one Datacenter/AZ per region. So you either risk a single region failure taking you down, or you lose performance from transferring data to another location.
So they could have had 100% redundant systems at OVH and still be under half the cost of a traditional "cloud" provider?
I would look at architecture and operations first. Their "main" node went down, and they did not have a way they could just bring another instance of it online fast on a fresh OVH machine (typically provisioned in a few minutes, assuming they had no hot standby). If the same happened to their "main" VM at a "hyperscaler" , I would guess they also would have been up the same creek. It is not the difference between 120 and 600 seconds to provision a new machine that caused their 10 hrs downtime.
If you're doing VPSes, then maybe, as long as they're not under the same node. If it's dedicated servers, then probably.
But I think "redundancy" is more like a spectrum, rather than a binary thing. You can be more or less redundant, even within the same VPS if you'd like, but that of course be less redundant than hosting things across multiple data centers.
And it's cheap enough that you can have replicated setup across two different providers and still be cheaper than one expensive cloud provider.
While AWS is probably towards the safer end if you want to put all your eggs in one basket, people are still putting all their eggs in one basket if they have everything at AWS as well...
But that question remains the same whether you are renting bare metal or VMs. You can rent OVH servers located at different datacentres all over the globe, and their Cloud SLA has higher uptime guarantees than AWS (what that is worth depends on the value you place on an SLA ofc.)
> when you have an outage, your stuff is tied to physical servers and they can't easily be migrated
I don't see how that follows? Could you please explain?
I run my stuff on Hetzner physical servers. It's deployed/managed through ansible. I can deploy the same configuration on another Hetzner cluster (say, in a different country, which I actually do use for my staging cluster). I can also terraform a fully virtual cloud configuration and run the same ansible setup on that. Given that user data gets backed up regularly across locations, I don't see the problem you are describing?
This is a myth, created so cloud providers can sell more, and so those who overpay can feel better. I've been using dedicated servers since 2005, so for 20 years across different providers. I have machines at these providers with 1000-1300 days of uptime.
Same here, been running dedicated servers with OVH since 2009, if anything bare metal server are more stable than before. I just replaced a set of servers that was from 2018, I didn’t have any hardware problems during their 8 years of working under significant load. During that time I had 2 or 3 power outages, a few more network outages. Usually problems come in a cluster. I had a few years that I had nothing to report, 100% uptime. Dedicated are nice, but I guess it scares people. Hetzner use lower hardware quality than OVH on some of their offerings, so your experience may vary. One of the most important thing is to check that your server use datacenter SSD/HDD with ECC ram, it saves you a lot of problems.
Most of them run Debian (some have Windows VMs running on those Debian machines), while a minority use Ubuntu. I reboot them once every few years when I upgrade the OS, kernel, or migrate to newer machine types.
I run most of the workloads in containers, but there are also some VMs (mostly Windows) and some workloads use Firecracker micro vms in containers.
A small number of machines are rebooted more often because they occasionally need new kernel features, and their workloads aren't VM friendly, so they run on bare metal.
My experience is exactly the opposite. None of the cloud vendors are actually resilient, every single one of them have had major global outages. And when it happens you've got no influence on how fast it gets fixed. The only way of building a truly resilient infrastructure eith cloud vendors is mirroring across vendors. But it happens to be easier to mirror a private cloud between e.g. Hetzner and OVH than maintaining parallel setups in Azure and AWS.
This is a very good point but even with dedicated servers it's doable to build a resilient HA architecture.
OVH offers a managed kubernetes solution which for a team experienced with Kubernetes and/or already using containers would be a fairly straightforward way to get a solid HA setup up and running. Kubernetes has its downsides and complexity but in general it does handle hardware failures very well.
A good example is a the big lichess outage from last year [1]. Lichess is a non-profit, and also must serve a huge user base. Given their financials, they have to go the cheap dedicated server route (they host on OVH). They publish an Excel sheet somewhere with every resources they use to run the services and last year, I had fun calculating how much it would cost them if they were using an hyperscaler cloud offering instead. I don't remember exactly but it was 5 or 6x the price they currently pay OVH.
The downside, is that when you have an outage, your stuff is tied to physical servers and they can't easily be migrated, when cloud provider on the opposite can easily move around your workload. In the case of Lichess outage, it was some network device they had no control of that went bad, and lichess was down until OVH could fix it, that is many hours.
So, yes you get a great deal, but for a lot of businesses, uptime is more important than cost optimization and the physicality of dedicated servers is actually a serious liability.
[1]: https://lichess.org/@/Lichess/blog/post-mortem-of-our-longes...