Use vanilla products (K8S, "RDS", ...) so you can always dangle the threat of going on-prem or to another cloud provider.
Once you use Lambdas and other deep AWS product, you're locked in.
Once you use Lambdas and other deep AWS product, you're locked in.
Each provider within TF is tied tightly with the cloud provider.
Standard Disclaimer: yes I work at AWS and I do IAC with the CloudFormation or TF depends on what the customer wants
You save a lot of time and resources just using SQS instead of setting up your own servers. And the lock-in is not as problematic as Lambdas etc.
It doesn't require much code change to move to an alternative message queue (RabbitMQ etc.), so the lock-in is minimal. In my projects, I just need to change a few lines of configuration code and the actual message processing code should remain the same.
There’s generally very little point using the cloud if you’re not going to use cloud services.
Especially if you need a couple of racks things get complicated quickly. If you have never racked machines for a week you probably underestimate the amount of time it takes before you can actually run workloads on them.
And if you need a 1000 machines for x years which otherwise you'd put in your own data center you can absolutely go to AWS, GCP, OCI and bid them down against each other.
“massive” is an overstatement. “some” would be more accurate. And your example is correct, which is why I said “generally” in my post rather than “always” or “never”. In fact my current project is using AWS in exactly the way you described.
> If you have never racked machines
I have done. Many times. :)
And there are ways you can get machines at short notice without going down the AWS route. In fact it’s actually pretty easy and used to be really common. I used to use all sorts of different VPS and dedicated hosting for personal projects (while managing on prem stuff professionally). These days a lot of those bigger companies are now just wrapped up as part of “cloud services” but there are still some smaller providers out there doing things old school. In fact I have a few GFX cards for machine learning in a London data centre managed by one such independent provider.
> And if you need a 1000 machines for x years which otherwise you'd put in your own data center you can absolutely go to AWS, GCP, OCI and bid them down against each other.
You’re now talking about cost optimisation and commitment-based discounts, which is a whole other tangent.
Anyways, my point wasn’t that the cloud is more or less expensive than on prem. it was that designing your cloud architecture to be cloud agnostic is more expensive and rarely worthwhile compared with making use of the services offered by whichever cloud provider you’re using.
But then again, migrating from one cloud to another is always a heavy lift no matter how it’s constructed.