Comment by mooreds
4 hours ago
> For personal projects I end up avoiding AWS and instead prefer things like the Backblaze S3-compatible object storage, Vultr for VMs, and so on just to avoid the power user features that will only get in the way.
The author wrote an article about this too: https://www.theregister.com/2025/11/04/aws_genz_misery_nope/
> With that, I am curious how people who do not have an enterprise-size team to manage their AWS infrastructure navigate their offerings.
I've been a startup CTO that used selected AWS infra (s3 buckets, RDS) along with an easier PaaS solution (Heroku, in my case). So I think the answer to your question is: using some of the managed services, which are rock solid, and using easier solutions for compute or some of the more complex AWS services.
I know folks who started similarly, but then moved to AWS fully when it made business sense (in one case, because of HIPAA regulations and the cost difference between AWS and Heroku for the BAA).
That article pretty much describes every experience I have had with AWS. But worse was when I tried to use the free tier of Oracle Cloud to see what that is all about. Oh my god what a mess. I am not new to a lot of these things. I know how to configure TLS, boot a Linux or BSD box when a hard drive fails, how to set up proper subnets and firewalls, hell I have written network services using raw IP packets that I crafted directly in C. But figuring out how to work the Oracle Cloud UI was beyond me.
It almost feels like if you were not there to see these interfaces in their infancy and didn’t grow up with them then you will not get their current much more complex form.
Back in 2008 I worked for an organization that relied heavily on an IBM mainframe and employed a department of people who managed it and wrote software for it. The divide between those folks and those who grew up on Linux/BSD was so solid that if someone asked me to switch teams I honestly wouldn’t even know where to start learning. This is kind of what this feels like.
To be fair I have successfully deployed multiple things with AWS. But it is always by mostly using things like EC2, Route 53, S3, and sometimes CloudFront. Tried their app engine and container runner and went back to running Docker on a Linux VM as a saner solution.
The problem is these small customers never drive enough sales to bother with—you're better off investing in a feature for a large customer. And by the time small customers get large enough to need things like complex permissioning, they've outgrown Heroku and will be onboarding anyway. Giving startups credits really might be the most effective way to handle rough edges for small shops.
As a startup, I'd probably bite the bullet of one-time setup pain for a database, blob store, load balancer, and service hosting at a major cloud provider because those systems will be rock-solid with well-understood APIs. Full disclosure: I work for a major cloud provider.
I used Backblaze and strongly regretted it.
Wonky bandwidth limits and throttling are my main problem, but also had some issues with login at one point which apparently wasn't unique to me. Would never trust it for anything mission critical after that.
The nice thing about S3 is even if you screw up your usage patterns, you can pay/engineer your way out guaranteed. You can slurp up as much data as you want as often as you want and it may not be cheap, but it will work and it can be made extremely fast.
I'm coming to find that's not universal for these S3 compatible services. Really scary to build a business knowing that.