[slovenlyrobot]

Since starting to use Google Cloud for bits and pieces I've come to appreciate the AWS UI approach much more than previously. All those little spartan pockets of UI means nothing gets overengineered, the tools feel more like a quick Intranet web app (and generally load as quickly!) than anything else

Meanwhile over in GCloud, almost /any/ operation whatsoever will spam you with an endless series of progress meters, meaningless notification popups, laptop CPU fans on, 3-4 second delays to refresh the page (because most of their pages lack a refresh button), etc., and the experience is uniform regardless of whatever tool you're using.

The uniform design itself was clearly done by a UI design team with little experience of the workflows involved during a typical day. For example, requiring 2 clicks and at least one (two?) slow RPCs to edit some detail of a VM, with the first click involving the instance name with any 'show details' button completely absent from the primary actions bar along the top. The right-hand properties bar in GCloud is also AFAIK 100% useless. I've yet to see any subsection that made heavy use of it

Underengineering beats massive overengineering? Something like that. Either way, the GCloud UI definitely pushes me to AWS for quick tasks when a choice is available, because the GCloud UI is the antithesis of quick

[Legogris]

Wow, I have the complete opposite experience. Monitoring, metrics graphs and logs are just miles ahead in Google IMO. It's so much easier for getting visibility.

Do you really prefer Cloudwatch to Stackdriver? How about having a Lambda being triggered both on SNS messages and HTTP requests (setting up a proxy) and having that Lambda deployed with a CD pipeline - compared to doing the same with Cloud Functions?

But I guess it also really boils down to which products you make the most use how, how and your scale. Clearly we have different preferences.

I guess I am not seeing the bad parts you do because 1) Apart from DNS and some IAM, most infra changes are done from Terraform or CLI and 2) I have pretty high-end workstation.

[slovenlyrobot]

A series of loaded questions does not convince anyone :) With large accounts all of these tools start to break down, by that point it's much easier to work from the command line than, say, navigating GCloud's single global view of VMs which becomes actively harmful in this case. I run 1500 instance load tests on GCloud quite regularly, so this is not some imaginary problem, and genuinely large accounts can easily grow to 10x that

I'll always prefer the ability to quickly hit refresh than waiting 4 seconds because I made the mistake of ctrl-clicking a link, and now a new tab is 'booting'. But I guess this preference depends on how quickly one expects to be able to get their job done

[Legogris]

I was actually genuinely curious, I guess ^^

Oh yeah, the inconsistency in which links open in new tabs by default and which you can ctrl-click and not is a bit frustrating in GCP for sure.

[timc3]

Total opposite experience. AWS UI is slow, workflow is terrible. Google cloud is much better.

Mind you once you get to a certain point using the APIs is better.

[sdan]

Agree. Having used both GCP and AWS, I like GCP's UI a lot more than AWS's huge dropdowns and constant usage of the search bar.

[mlthoughts2018]

I have to disagree. After using AWS & GCP, I find AWS “stays out of my way” much much better and has much better documentation. There are weird corners of GCP, like GCS “interop” mode and lack of full compatibility with S3 APIs that feel basically like Google is using dark patterns.

AWS is head & shoulders the better cloud provider. Google is just cheaper.

[jjoonathan]

I love when AWS console stays out of my way so much that I have to pick through its backend HTTP requests in the browser developer tools to figure out why it isn't working even though the UI shows no errors or perhaps even reports success. This is a regular occurrence.

My absolute favorite is when AWS console stays out of my way in a particular manner that hides expensive resources, with bugs in the per-resource console, the cost explorer, and the notifications systems conspiring with each other to deliver a lovely surprise at the end of the month. It's amazing what bugs can do when they work together!

[lostmyoldone]

Yeah, or when you try to open an account get stuck in limbo where you can't really access anything in the console, and suddenly you start getting bills for those Jungle Disk disk S3 buckets you forgot about years ago!

Somehow my new AWS account got linked back to those old S3 buckets, and something went terribly wrong. I really need to try to get that fixed somehow, while it's only $7 or so a month, it is still $7 for exactly nothing. Since I to my knowt haven't had an account for the better part of a decade, I was quite surprised that it was actually still retaining those backup buckets, but also that I started getting bills for the bucket. I believe the terms when JD was liquidated was that the buckets would remain, but be free of charge. Well, they were free until I created a new account.

Last time I tried to shut them down I think I actually managed to set up CLI access, but then I got sidetracked by actual work. First time I tried, AWS wouldn't - figuratively - even show me the time, so getting CLI access wasn't even possible.

[mlthoughts2018]

It sounds to me like you’re describing GCP, based on my experiences.

[heliodor]

Just because something is better does not make it good.