Which in itself is a massive selling feature. Deploying a simple app on AWS is about as straight forward as deploying Active Directory in the early 2000s. It is so bad Amazon even launched the Lightsail product line to address the problem.
AWS now is basically "you don't know how to run a database, or a KV store, or a network, etc. so here it all is as an API with a markup." Some companies still have talent that knows how to run things efficiently and just need CPU, RAM, and storage.
It's not at all a selling point. You can do the same on AWS, get an EC2 and install everything there, that's it.
The thing is that when you will need to leverage the other services of AWS, they will be there, while you will need to migrate if you're on one of those server providers.
I never understood the complaints of complexity about AWS. You can easily just use EC2 and maybe S3 and ignore the other 100 icons.
I think, in practice, what folks are complaining about is that AWS has the worst UI known to man. In which case, you can use GCP for their nice UI and still get the benefits of a hyperscaler.
I don’t mean to say that the simpler services have no place in the market but it’s pretty narrow. The fact DO and Linode offer things like Kubernetes now is evidence of that.
Also hyperscaling is a bit of a lie. You have a trillion invisible limits built into your account, you have to talk to support to raise, if you try starting up instances by the hundreds, you'll be throttled.
There's zero transparency on what types of instances are available and how in-demand they are, meaning trying to scale up capacity 'on demand' is a fool's errand.
Even worse with Azure. They frequently freak out if you try to provision more than 1-2 instances per week. I have encountered this issue multiple times.
"AWS has the worst UI known to man, except for Azure and GCP".
It still amazes me that when you add members to AAD groups the interface pops up and tells you 'change may not be displayed yet'. It's like, YOU HAVE ONE JOB.
I honestly would not trust a developer who finds AWS too confusing. At least 80% of the "complexity" on AWS is to keep you from introducing catastrophic security vulnerabilities.
When a developer starts whining they can't figure out how to use an IAM role, it usually means they want to run their entire application as a root user whose SSH key they have sitting in Dropbox.
This is called Stockholm Syndrome and there are people who can help.
AWS UI is filled with manual and tedious bullshit to do every day things. It really doesn't have to be this way, making things annoying to configure correctly leads to mistakes or lazy overly broad permissions.
If they weren't the first they wouldn't have made it.
You’re not describing a poor developer. You’re describing a poor AWS developer.
There are plenty of excellent, security conscience, etc developers who take a look at IAM or any of the other dozens of sticky AWS traps and run away screaming.
As someone who spent five years as a consultant for two AWS pro partners, I agree with this anecdote. Day one of any gig when I started building with my clients was training on IAM and resource policies.
> I honestly would not trust a developer who finds AWS too confusing.
So you would not trust most developers?
Even a lot of Sysadmins i.e. Devops find it confusing.
A lot of people just pretend it is fine by being ignorant and not learn any of it. They would follow 1-2 processes that works and never do anything else. You ask them why or how it works and you can't even get a real answer out of it.
Unless I actually need one of the services that's not available on a tier 2 provider, I would personally never want to use any of the hyperscaler services if all you need are covered by the services offered by the tier 2 providers.
And this is mostly for security reasons: I'm not familiar with Azure but AWS and GCP setup is very complex and it's very easy to make a mistake that would impact the security of your infrastructure. Tier 2 providers like DO, Scaleway, Linode before Akamai, etc generally offer products whose setup you can fully understand and that helps a lot in giving you the confidence that you didn't mess up and shot yourself in the foot without knowing it.
You're right that the UI is crap, but there's also another dynamic when you work in a team: some people may not be aware of complexity, and can create overengineered monsters just because they want to try the fancy stuff.
That might suck, they leave their job, and you're stuck maintaining the monster.
Was looking for this comment. On the bright side, it's easy to get lulled into a sense of complacency when you're running everything in the cloud, and OVH's massive data center fire had me double check all of our backup and disaster recovery processes (on a different provider) to ensure they were geographically distributed.
But they might be one of the greenest. OVH had to make some trade-off compared to other providers to remain as cheap as they are and while it certainly has an impact on their reliability and security levels, a lot of them are actually positive as far a sustainability goes (use of recycled containers, the use of wood in the DC which was decried when their DC burned down etc).
To put it another way: a DC in a wooden shack connected to a low carbon grid (France) surely has bad security but a low environmental footprint too.
