I actually thought the opposite. If you are small you want to move quickly and spend as much energy as possible on developing the product not managing infrastructure. There is a lot involved, OS updates, certificate management, secrets, multiple tiers of backups, but also monitoring and alerts, etc.
If you have a lot of customers and server cost on the other hand, it makes sense to employ a dev ops team, that can focus on optimizing costs.
Technically he is suggesting renting a dedicated server hosted in a Hetzner remote data center. So in some ways it's halfway between "cloud" and "on-premises" hosting.
Doesn’t fully convince me as having a single physical server is very different than a managed DB. If you have a disk failure you will be down until that is resolved. Which is in fact the exact scenario that happened to a client of mine and which led me to migrate to the cloud
Unless I missed something this is a VPS in a cloud, so really its not about replacing hardware to get back up.
There's certainly an interesting problem of scale and redundancy where I think you need 3+ half full racks to really be efficiently running in DCs and then have to consider how their markup relates to your inefficiency if you are running a lot less.
I actually thought the opposite. If you are small you want to move quickly and spend as much energy as possible on developing the product not managing infrastructure. There is a lot involved, OS updates, certificate management, secrets, multiple tiers of backups, but also monitoring and alerts, etc. If you have a lot of customers and server cost on the other hand, it makes sense to employ a dev ops team, that can focus on optimizing costs.
Oh, it’s time for the cycle to flip back to on-premises servers?
Technically he is suggesting renting a dedicated server hosted in a Hetzner remote data center. So in some ways it's halfway between "cloud" and "on-premises" hosting.
He's clear about going for smaller use-cases, not enterprise-scale.
Doesn’t fully convince me as having a single physical server is very different than a managed DB. If you have a disk failure you will be down until that is resolved. Which is in fact the exact scenario that happened to a client of mine and which led me to migrate to the cloud
Unless I missed something this is a VPS in a cloud, so really its not about replacing hardware to get back up.
There's certainly an interesting problem of scale and redundancy where I think you need 3+ half full racks to really be efficiently running in DCs and then have to consider how their markup relates to your inefficiency if you are running a lot less.
With serious users, I would feel compelled to warn about the outage risks I were taking.
Honestly, us-east-1 has gone down more times in the past 5 years than my VPS has.
I agree that their "availability zone" story isn't solving the problem, so a quorum should span regions rather than just us-east-1a through 1e.