Why the hell are companies going to AWS and Azure instead of GCP??? I mean for Azure I understand compliance is a little easier with HIPAA and similar things but seriously.

GCP is so transparent about everything and it's simple for everyone.

GCP has:

- lower market share, smaller community
- fewer humans skilled in it to hire
- less robust integration

@SortOfTested true, Azure has the beautiful AzureAD integration with IAM.

I can justify Azure and GCP just AWS seems so crappy in comparison.

Was it because they were one of the first to even start public cloud?

Also, a very complicated billing logic. Making it Hard to compare between the platforms.

@magicMirror I think all cloud providers have their billing problems.

- Azure has 3+ types of instances for "serverless" functions. Everywhere from consumption which isn't really consumption to dedicated plans.

- GCP has gotten a lot better over the past few months making it a little more transparent but their panel still sucks.

- Never spent enough money on AWS to mess with billing reports so that's past me. If their panel looks anything like it used to though I think it would give me an aneurysm.

Cause of how people think.
ex. AWS was first and have biggest availability of people who know it so lots of managers correlate cloud with AWS. There is also probably more mature tools for AWS migration and list goes on cause of being first.
Also AWS invest lots into consulting and spending lots of money giving free credits.

Azure is obvious for windows / office ecosystem and corporate relationships with it.

GCP is mostly used for machine learning cause of google is associated with it.

@n3xus
It's not really that simple. AWS is the best of breed because it has significantly more diverse offerings, and puts the fewest abstractions between the underlying technologies and the APIs that expose them to you.

It also has the closest to scale to zero on all its mainline offerings, and has myriad cost management options including reserved instances and bid infrastructure.

If you're just building a small app, nearly anything will do. If you need to run a business and get max control, visibility and insight, that's what AWS offers in an unopinionated, os-agnostic fashion.

@vane Googles tools are fabulous for building apps from the ground up. Look at snapchat and their success with GCP.

I think AWS and Azure are better solutions for smaller businesses and people looking for integration with non cloud-native applications.

@SortOfTested what scale to zero offerings do they have available? GCP can spin up a kubernetes instance when an HTTP request hits a load balancer or even a TCP request.

GCP also has cloud functions which spin up crazy fast and scale to zero as soon as they aren't in use.

Also thank you everyone for sharing :) I only know so much and I'm somewhat new to the industry.

@n3xus yeah maybe but are they consistent ?
I remember old famous ex amazon employee rant on google services.

Also google is known for shutting down their products.

Rant archive link:
https://gist.github.com/chitchcock/...

@n3xus
We're entirely scale to zero on eks/ECS on dotnet core 3.1 and micronaut.

EMR jobs can be scale to zero, and robust FaaS via lambda are other examples where your requisition can be based on spot instances.

@vane
That guy's memory fails him a lot. I did the exact opposite of him (n years Google, n years Amazon). The mandate he's talking about was about 5 years ago (he's embellishing his Google experience by that metric) in the utf-8 all the things+https is the canonical protocol initiative. That was establishing a baseline protocol across the company, not limiting it to one thing. He obvious left before they dumped Oracle, also.

It sounds like he was in the .com org. That is kind of a painful shitshow, but it's almost 1500 teams building a single website. The fact it works at all is a testament to the skill present. AWS's startup model worked very well and produced a litany of products that lead the industry, I enjoyed it thoroughly. I also enjoyed google, but I can't say it's categorically better. I've seen people get hired into Google I knew had no skills (political hires, nepotism hires, etc) far more frequently than ever happened at AWS.

@vane I read the rant and it seems to be pretty much: "start with a platform and build on it. Don't build the platform after" which I agree with whole heartedly.

Any successful modern technology works flawlessly like Snapchat or Instagram.

Look at how garbage Facebook is, they started with LAMP and instead of rebuilding, they keep pushing the same garbage.

@SortOfTested I also look at things like how much Google developers contribute to community projects like kubernetes.

Amazon is a leach and contributes nothing back to the OS community despite using a ton of it in their cloud.

@n3xus
I wouldn't be so sure. I contributed a ton to even Google initiatives, but I did it on my personal account. I have half a dozen pushes to elasticsearch and dozens to angular, and bits and bobs around problems I had with tools.

What Amazon doesn't do is try to buy controlling interest in Open source steering committees, like say, Oracle, IBM and Google.

@SortOfTested that's a view I didn't even consider. I am friends with a few Google engineers and from what I'm told, they actually have teams only working on open source projects that Google doesn't even have ownership of.

I guess it could be seen as good-will but could also be seen as gaining controlling interest. Good insight :).

I'll give AWS a shot and consider getting a certification in it. I just find that it seems more like one of those companies that pushes as many services as possible and never bothers updating them or innovating.

@n3xus there is not such thing as modern technology, all technology that is developed is obsolete in a moment it’s shipped to customers or when twice as many customers start to use it.

There are modern problems and as far as I know all of them are related to data processing, gathering and storage.

That is what you can’t see - amount of data companies are pushing to cloud and I’m amazed it still works and we didn’t have any serious problems with all of those data centers and cables spread around the world. All of those hard drives.
All of it is to beautiful to be true.

In me experience, Azure & AWS have a lot lot more to offer than GCP. Also, GCP charges 2x, sometimes even 3x the amount that too with a lot of limitations (at least for the services I was looking into)

@DangerousDev I think a lot of people get things confused with GCP.

U like AWS or Azure they really could care less about existing apps. They want all the apps on the platform to be purpose built in the way they expected them to be built.

This is nice if you're a dev and you do it right the first time, but as soon as you are porting existing infrastructure or code over it becomes a logistical nightmare.

I would say from all my platform research last night, Google Cloud Platform has one of the most ideal platforms for building an app from scratch so long as it has no super-special/odd requirements. As soon as you need something special on GCP is when the costs skyrocket.

@n3xus Actually in my case we were building things from scratch. Our architecture is completely cloud native with main server work completely on Azure functions (GC Functions costs 2x this). Also, it is an IoT based project and Azure & AWS had a far more robust toolset...!

@DangerousDev azure functions from my experience aren't serverless. You can go consumption based but your app won't respond to http requests while it's spinning up.

@n3xus Azure functions are serverless! Consumption plan is just a payment model...

@n3xus And if you're referring to cold starts, the happen in all three: Azure Functions, Google cloud functions and Lambda..!

@DangerousDev yes but Google functions spin up in less than 5 seconds. Azure will just forget about your http request.

As for payment model, you are paying for A VM insurance unless you use consumption.

As for AWS I have no clue

@n3xus No it doesn't! We are using it in prod! The cold start time is not even noticeable! And it DOES Not forget about the request! Are u sure you didn't do anything wrong while trying!?

@n3xus For the payment thing, the consumption model will fireup new instance whenever u need it.

The Premium mode is a hybrid thing, with some instances already hot and running while the others are fired up whenever u need them.

Both kinds of models are really important depending on the nature of service and scale!

Okay, so give u some numbers here, for out dotnet core azure functions, it takes around 2-3 seconds for a cold start!

@DangerousDev ok so I made a nodejs app with a consumption http app. It took up to 10 minutes at times just to spin up the instance. We had to switch to a development plan to make it stay hot.

Http requests would timeout after about 10 minutes and you would have to refresh, at which point the function has gone to sleep again.

@n3xus Wow, 10 mins is long! I guess there was something wrong with the app (or maybe how the dependencies were handled)... They do have a default timeout of 600 seconds.

I have tried smaller things with node functions, but never faced any issues with startups.

BTW, which runtime version were you running on?

@DangerousDev I hope the app wasn't doing an npm install everytime it was spun... 😂... No clue what else could've taken so long... 🤔