[lobster_johnson]

These days, GCP completely blows away AWS in every area where they overlap, especially VMs, networking and cloud storage.

GCP is generally a lot more consistent and modern, having benefited from a clean-slate design and Google's expertise in designing huge continent-spanning infrastructure solutions. Some examples:

* AWS Glacier looks pretty nifty until you realize that Google's equivalent, the coldline storage class, still gives you millisecond latencies (Glacier's is usually measured in hours), at a much lower price point.

* GCP is designed to automatically migrate VMs between physical hosts, with zero application downtime. This in turn is possible thanks to their rather amazing network stack, a completely transparent, encrypted L4 SDN. AWS, meanwhile, suffers on the awkward mess that is AWS VPC (unless one is still stuck on "legacy EC2").

* Google gets things like consolidated billing correct from the start. Other minor things like IAM management and the command-line toolchain are a breath of fresh air.

Not to say that it's all rosy, of course. GCP has its own set of issues, like any product offering, though no major ones come to mind. But it's clear that some of their services (the StackDriver tools comes to mind) are not up to their usual quality standards.

Of course, Google doesn't have counterparts to all of Amazon's offerings, but then AWS has a lot of odd products (much of it targeting enterprises with legacy infrastructure). Google's focus on Kubernetes and containerization means that some of the deployment-oriented services (CloudFormation, Elastic Beanstalk and so on) are less relevant, and for now Google seems to rely on the community/third parties to come up with their own solutions, which I think is a good call at the moment.

[runeks]

Another cool thing I've experienced with Google Cloud: a while ago I had a VM running, with the default 7.5 GB RAM for a 2 CPU VM. Then when I came back some days later, I got a message saying that the VM had only been using ~3GB of RAM, and that I could save money by reducing the VM's RAM, by clicking "this button". Then I clicked the button, the VM's RAM was reduced and, if I recall correctly, it didn't even require a reboot.

That's not only technically impressive, it's also really good customer service.

[lobster_johnson]

You can't resize running VMs — Google is good, but not that good. You can stop, resize, and start again, which is still better than AWS (unless they support changing the instance type now). I suspect that the button you talk about did that.