Ask HN: Which cloud behemoth is best for individual learning?

[mardnart]

I want to get familiar and comfy with cloud concepts like auto scaling, serverless functions, and all that. Which cloud platform is the best all around for learning: AWS, Azure, or GCP? I’d like to take into consideration the cost at low scale (spending double digits per month is fine, probably), the quality of documentation and tutorials, and the breadth of services. How do the three cloud providers stack up for hobbyists, students, and such?

[funkjunky]

I'll say this in GCP's favor: it has an extremely generous free tier and a $300 credit for signing up, a ton of fantastic tutorials and walkthroughs at https://codelabs.developers.google.com, and they seem to be at the forefront of most major R&D advances in cloud computing over the last couple of decades, so you'll probably see new things there first (managed kubernetes, Spanner, and a wealth of fancy machine learning shit just to start).

I also recommend going through the "Google cloud architect" and "data engineering on Google cloud" courses on Coursera: https://www.coursera.org/googlecloud.

The docs have gotten much better over the years, but you have to put in the time to really dive into them. They are essential reading, so put in the time.

Also the developer tools, (web dashboards, CLI tools, APIs, code libraries, etc.) seem to be light years better than what I've experienced with AWS and Azure.

Disclaimer: I used to work Platinum support for Google cloud. Now I'm an SRE at a company that uses AWS, and I play around with Azure here and there to see what's what

Whatever you decide to do, I recommend coming up with one idea for an application that can start simple but grow into something complex, start from the bottom with a simple deployment, and add pieces as you learn them so you can understand how everything fits together in the grand scheme of things. I picked election/voting software, which is good because it starts at just simple vote counting over an API, then grew to teach me auto scaling, data pipelines, machine learning analysis, etc.

[weber111]

My networking lab class (at a top5 school) used AWS, with the reasons that it's more widely used and better documented than GCP or Azure.

Tip: You get more free credit (by like 50% iirc) if you sign up for AWS Educate via the github student developer pack than if you sign up directly

[swaggyBoatswain]

Ive been told by many developers firebase does have the lowest learning curve and friction, gets the job done for smaller projects.

Plus AWS's UX is just plain awful and backwards.

Azure has the best jupyter notebook support for free too

[neilsimp1 → swaggyBoatswain]

Eh, Azure's UI isn't much better. I mean, I really do like Azure for the most part, but I feel like every time I go into the Azure portal online (every few months) I need to Google how to find something because it got moved around on me. And half of the results I find are out of date.

[funkjunky → swaggyBoatswain]

Google has free collaborative jupyter notebooks at https://colab.research.google. They also have Datalab in GCP

[ro_sharp → swaggyBoatswain]

If you’re using the AWS console on the reg, you’re doing it wrong (but yes, it’s a mess)

[AtomicOrbital]

Invest several days teaching yourself each of them (azure/gcp/aws) ... spin up a VPS and execute some code then build a serverless service ... I tried each two years ago and felt gcp was the most enduser friendly ... do not limit yourself to using those explore open source github repos which provide a serverless framework then execute you code on a much cheaper VPS from places like https://www.ovh.com/world/vps/ where you can get a decent 8 gigs of ram box for $14/month

Create a toy domain at domains.google.com which you can use anywhere your compute runs ... write a server then wrap it with a container and spin up instances of it to get the feel of microservices ... pickup the language golang and start using some of the hundreds of API's available

https://www.programmableweb.com/category/all/apis

Throw together several servers just to stream data between them using different protocols like DASH-MPEG or websockets

Become a master of visualization using D3.js or WebGL to enable you to render dynamic interactive 3D browser friendly graphics

[scarface74]

Go with the old maxim that “no one ever got fired for buying IBM”

The most popular and marketable ones are AWS and Azure. Azure is more popular in large enterprises.

[mr_toad → scarface74]

> Azure is more popular in large enterprises

The Azure-branded swag is certainly popular in large enterprises.

[kaishiro]

Ultimately I think you'll learn a lot from any of the big three, but in my personal opinion Firebase Functions (really just a thin wrapper around Google Cloud Functions) will get you parsing request/response with the least amount of friction. So if you really just want to learn the ins and outs of cloud funcs/async models it'll get you in to code really quickly.

e.g. -

  npm i -g firebase-cli  
  firebase login  
  firebase init  
  // Modify index.js  
  firebase deploy --only functions

[auganov]

For quite a while AWS and cloud were basically synonymous. It took years for others to get in the game. And they're still playing captchup. Because of this quite a lot of 3rd party software targets only the AWS. I mostly hear about Azure and GCP in terms of discounts, deals and the likes.