I work for a small company owned by a huge company. We are entirely independent except for purchasing, IT, and budget approval. We run our CI on AWS, and it’s slow and flaky for a variety of reasons (compiling large c++ projects combined with instance type pressure). It’s also expensive.
We planned a migration to move from 4OD instances to one on prem machine and we guessed we’d save $1000/mo, our builds would be faster and we’d have less failures due to capacity issues. We even had a spare workstation and a rack in the office that so the capex was 0.
I plugged the machine into the rack and no internet connectivity. Put in an IT ticket which took 2 days for a reply, only to be told that this was an unauthorised machine and needed to be imaged by IT. The back and forth took 4 weeks, multiple meetings and multiple approvals. My guess is that 4 people spent probably 10 hours arguing whether we should do this or not.
On AWS I can write a python script and have a running windows instance in 15 minutes.
This is the root success of aws, it lets internal teams bypass sysadmin departments.
Yup! Exact reason I saw companies migrate to AWS. Basically incompetent sysadmins who could barely manage a toaster let alone a server rack. I have this memory of one "sysadmin" who was tasked with expanding the server farm, asking me for help for what to order from Amazon.
I think it's the common reason why companies end up outsourcing technical work, whether it's to cloud or consulting companies or whatever. Leadership hasn't built a strong engineering culture (which includes hiring bar, how you promo people, etc.). This means companies are forced to spend 100x by outsourcing everything to get anything working, and AWS is one convenient way for that.
The same story applies for software. If I want to buy a license of X for someone, I have to go through procurement, and it takes weeks even for <$50 purchases. Yet if its on the AWS marketplace it’s pre approved as long is doesn’t breach the AWS budget.
Working around official IT was certainly a significant factor early on. I'm less convinced it is nearly as big a driver (or a downside depending on your perspective) today.
It depends on organization size, just my anecdotal example, I would say the moment IT department becomes own island (for example: can totally ignore requests, with excuses staff overbooked/we need extra planning/6 months extra meetings. Or even worse - process request,but up to point where it can show for upper management and blame you for wasting resources) - you can go full cloud, at least there it is possible get something working in reasonable time.
Especially considering that outside of startups (where approval would be fast with or without cloud), virtual infrastructure also got its own bureaucratic process.
A lot of people forget that, when server virtualization was still gaining momentum in a lot of circles, it wasn't uncommon at less technically savvy customers--say a regional bank at the time--to be told that it might take 2 months to provision a new server.
I don't think anyone is forgetting that in this thread, as there's dozens of answers mentioning this.
But as an example: It took about 3 months to provision an AWS server in a recent company I consulted for due to their own bureaucracy and ineptitude of the Ops team.
On the other hand, when I needed a few CI servers for a startup I worked at, I just collected them from AppleStore during lunch hour.
Now this above is what people are "forgetting" and don't want to listen to.
For us the problem is every device that gets plugged into our network is disabled by default, IT need to enable the port and they'll only enable it on machines that they've imaged.
But because AWS isn't in the office, it's fine. We could probably use Hetzner or OVH, but then we have to go through procurement which is as much of as hassle as going through IT.
Now there's two dozen posts with the same example repeated.