Sadly ISP's won't support this without an arm and leg. Small businesses can jump onto Cloud, but no way they can self host. Even getting a email provider that gets past spam is really hard.
>Small businesses can jump onto Cloud, but no way they can self host
They don't need to choose between 100% Cloud or 100% self hosted. There is a middle ground where they can choose one of the many providers of hosted computing (like Hetzner). They can save a significant amount of money.
> Even getting a email provider that gets past spam is really hard
Mail can easily be solved by using a third party provider like Postmark, instead of using SNS or whatever service your cloud provides. I have experience with both, and there was not much difference in the time I spent to integrate them. It's also much easier to test from your local machine compared to cloud services.
> It's really hard to build a system for the cloud that can run on bare metal unless you make that a goal of the project
We see a trend today that many are using Kubernetes to build their platform, and that can run anywhere.
Unless you go all in on Lambdas and step functions, you are usually not that dependant on cloud (or K8s for that matter).
My workplace has simple services that are built as Docker images running on K8s, connects to Postgresql databases and sends messages using SQS. Only SQS is specific to AWS, and can be easily replaced with a different message queue (we are using Spring abstractions).
> Especially when it requires more head count.
Quote from the article:
"""The cloud was not able to make things easier to a point where we could measure any productivity gains,” he says, noting his operations team has always been about the same size."""
You still need an ops team once you scale up from a small startup and you don't necessarily end up with lower costs for the team itself. And if you take the added cost of the cloud infrastructure, it should in theory get harder to justify for companies as they grow. But I guess "Nobody ever got fired for choosing Cloud"
As DHH himself says here: The cloud allows flexibility, which is essential when you are figuring out what you are doing, as in a startup. If you are large and stable and know what you want, then yes, you can go on-prem. But note that the cloud providers will negotiate significant discounts for large customers.
So the cloud is slow, expensive, complex and inflexible.
Better to run your own machines.
Sadly ISP's won't support this without an arm and leg. Small businesses can jump onto Cloud, but no way they can self host. Even getting a email provider that gets past spam is really hard.
>Small businesses can jump onto Cloud, but no way they can self host
They don't need to choose between 100% Cloud or 100% self hosted. There is a middle ground where they can choose one of the many providers of hosted computing (like Hetzner). They can save a significant amount of money.
> Even getting a email provider that gets past spam is really hard
Mail can easily be solved by using a third party provider like Postmark, instead of using SNS or whatever service your cloud provides. I have experience with both, and there was not much difference in the time I spent to integrate them. It's also much easier to test from your local machine compared to cloud services.
What’s old is new again.
The trick is to make money regardless of which way the hype train is going.
Related from December:
37Signals Left the Cloud
https://news.ycombinator.com/item?id=38705886
Leaving the Cloud: Cloud Computing Isn't for Everyone
https://news.ycombinator.com/item?id=38644550
It's really hard to build a system for the cloud that can run on bare metal unless you make that a goal of the project.
That can be a hard goal to sell to business. Especially when it requires more head count.
> It's really hard to build a system for the cloud that can run on bare metal unless you make that a goal of the project
We see a trend today that many are using Kubernetes to build their platform, and that can run anywhere.
Unless you go all in on Lambdas and step functions, you are usually not that dependant on cloud (or K8s for that matter).
My workplace has simple services that are built as Docker images running on K8s, connects to Postgresql databases and sends messages using SQS. Only SQS is specific to AWS, and can be easily replaced with a different message queue (we are using Spring abstractions).
> Especially when it requires more head count.
Quote from the article:
"""The cloud was not able to make things easier to a point where we could measure any productivity gains,” he says, noting his operations team has always been about the same size."""
You still need an ops team once you scale up from a small startup and you don't necessarily end up with lower costs for the team itself. And if you take the added cost of the cloud infrastructure, it should in theory get harder to justify for companies as they grow. But I guess "Nobody ever got fired for choosing Cloud"
Helm charts are much easier to distribute across teams than Cloudformation or Terraform modules, which is pretty crucial.
>> It's really hard to build a system for the cloud that can run on bare metal
That’s an overly broad statement and likely false for many workloads .
How do you factor in two 9s versus seven 9s of uptime?
As DHH himself says here: The cloud allows flexibility, which is essential when you are figuring out what you are doing, as in a startup. If you are large and stable and know what you want, then yes, you can go on-prem. But note that the cloud providers will negotiate significant discounts for large customers.