Now go read the aws documentation on an ALB. One page, this is how you create one and this is how you use it.
Azure on the other hand arrived late with a spectacular business plan. “You already pay me for software, cloud is part of that”. They won enterprise cloud in the blink of an eye.
Meanwhile Gcp wrestles with writing task oriented documentation, easy things are obscure or weird. (Compare security groups at aws with gcp, now look again. You didn’t get it the first time ) also if you’re designing a service there’ll be some critical thing that just doesn’t work. The old thing is deprecated the new thing is beta. (You fool! Don’t use beta things in production! Oh sorry there’s the non beta thing I can use? There isn’t one) or the feature you need to complete your task will never arrive because there’s a pissing march between gcp and k8s over who’s job it is to make the ingress work as desired.
Also, as others have said their support is garbage (I found a customer evangelist who was amazing if I could wait 3 days to get a response)
Next is poor support. I worked in a team that used DialogFlow enterprise for a voice app. The NLP engine was way way better than Alexa. But if you face an issue their enterprise support will tell you to post in StackOverflow !! As an example of why they don't get enterprise, DialogFlow didn't support mTLS for authentication and had no plans to build that feature.
AWS has the first mover advantage and enormous brand penetration - it's literally the "you can't get fired for choosing AWS" (instead your company might fold due to AWS bills, but you'll be clean).
Meanwhile Google seems to be much less known in cloud infrastructure space - people think of Google Apps/Workspaces, not GCP, and they seemed to have serious problem getting to clients. I know I was quite surprised by the available stuff and low prices when I did my first project on GCP back in 2016, and generally I've been pleasantly surprised except by their horrible billing support (Google can't into actually getting paid?)
They are in this funny situation where they constantly seem to be lacking fully blown virtuous cycle of adopters and supporting companies, despite having some of the best options on the market - but I think the real issue is that they are neither the first mover (AWS) nor "canonical windows source" as with Azure.
Some of that impression is leftover from the way App Engine handled things and the first impression that left, but even GCP documentation itself still seems to struggle outside of "Google approved" languages.
Disclaimer: I interviewed with GCP as a .NET developer with a lot of experience in .NET to try to improve the situation. I got a lot of nasty ego directed at me that interviewers didn't trust my technical expertise because it was in such a "gross" stack and didn't know how to interview me technically. I personally saw that as direct evidence for why I'd never use GCP.
It doesn't matter how good the services or pricing are. I dont trust them and I doubt I ever will.
Today, GCP's still better-positioned for small companies who need to move fast and are more price-sensitive, but they've done a good job catching up on the enterprise space, and having late-mover advantage has helped them avoid some of the footguns you see in AWS. There are still a lot of sharp corner cases in their services and documentation, and they're as bad as anyone else in Google when it comes to taking customer direction (it really helps to have direct contacts in the organization or be a large-scale implementation partner), but they're a perfectly valid cloud option with a lot of great services at relatively aggressive price points (and, arguably, if you're working in health or life sciences, they're actually a very good option compared to AWS or Azure thanks to their extensive healthcare API portfolio). Unfortunately for them, they're never going to be anything more than the fourth-largest global cloud provider, which isn't a bad place to be, but probably a bit humiliating for GOOG.
I see that hot swapping the identity of a node pool may be difficult, but it's weird that you can't change the firewall rules.
Overall I prefer the philosophy of AWS regions vs global infra. So many GCP outages were global. Also AWS seems very keen on customer feedback (if you are big enough at least).
