AWS vs Azure vs Google Cloud

In the second post of this compare/contrast/debate series, I present the biggest cloud platforms.

You're welcome to pick your favorite and debate it out, but comments comparing the major differences are very welcome. 🙏

Comments

[Andrew Brown 🇨🇦]

AWS > Azure > GCP

[Joseph Friedman]

Can you elaborate?

[Basti Ortiz]

Well, that was straightforward... 😂

[simonhaisz]

This certainly aligns with marketshare

[Dave Jacoby]

Accepting this as the Codd's honest truth, I would like to hear more explanation about why. Capability? Complexity? Cost?

[Raunak Ramakrishnan]

I have used GCP and AWS so far... I favor AWS slightly over GCP.

GCP

Pros:

• Cost
• Browser SSH for quick debugging (cloud shell)
• Instant resizing of disks, memory for the compute instances
• Cloud Pub/Sub is great
• Firebase is useful for prototyping

Cons:

• Support. They need to improve this and be more customer focused.
• Reliability - There have been a few incidents where their network was down. But they have been quite stable since.

AWS

Pros:

• Reliability - The best among the major cloud provider
• Support - They understand customer pain points very well. Even their free tier support is helpful.
• Documentation - Comprehensive and generally up to date
• They have never killed any existing service. Even deprecated ones are still around though customers are discouraged from using them.

Cons:

• Cost - AWS is generally more expensive than GCP
• Calculating costs - Sometimes I feel we need a PhD in Cloud Economics to calculate AWS bill especially with Lambda, API Gateway, Cloudwatch.
• Too many services with weird names or acronyms. Keeping track of them can become cumbersome.

[Vitor Amaral]

This here "Sometimes I feel we need a PhD in Cloud Economics to calculate AWS bill " is so true. 😂

[Thai Pangsakulyanont]

I use both AWS and GCP.

On GCP I use:

• Cloud Speech-To-Text. I don’t see any competitor that gets it more accurately than Google.
• Firebase Realtime Database. I don’t see anything that’s more simple to use.
• Cloud Run. I don’t see anything like this on AWS. Tried Fargate before but the last time I tried, for a large container, it times out booting up. I don’t see any real competitor for Cloud Run yet.

On AWS I use:

• Cloud9. I like it that I can choose from a whole catalog of EC2 instance types and pay just for the EC2 price. Cloud9 auto-shutdowns the EC2 instance when unused, so an $80/mo machine ended up costing me just $20/mo. I also use Cloud9 to develop projects running on GCP. I tried Google Cloud Shell but for me it’s not a good substitute.
• Lambda. The last time I benchmarked it (by invoking simultaneously thousands compute-intensive functions) I found Lambda to be the most stable.

[Geshan Manandhar]

Cloud Run is amazing.

[Filip Biterski]

I've just published an app on AWS with cloudfront distribution. What I can say is, even though many claim it's very complicated, it's surprisingly well documented. I registered a domain, uploaded files for a static site on S3, and setup a cloudfront https distribution in a day.

[Heinek Halwin]

Also the caching capabilities of the cloudfront network at their edge locations. 👌🏻

[Wernich ️]

azure is by far the best one i've used so far*

*i've only used azure. :D

we use azure devops at work for source control with git and last week i decided to give the extra bits a whirl, so now i'm busy reading up on pipelines and deployment and so on (having tons of fun with their learning paths). also, i committed to doing a showcase this friday about using azure pipelines to build and deploy a docker container when a PR succeeds...

[Andrew Vitale]

Nice! Similar situation here. My company has a Visual Studio Pro account for each dev, which comes with $50 worth of monthly credit in addition to the DevOps pipelines/artifacts. Just started playing around with Blob Storage and their CDN. It's quite nice, but I do prefer AWS a little bit. First love, I guess 😍

[Fayaz Ahmed]

I use Digital Ocean for my small projects & I even recommend it to my clients.

[Sean Allin Newell]

I'm swimmin' in the droplets.

[Pacharapol Withayasakpunt]

VM aka DO Droplet is not for everything. Kubernetes in DO should be similar to other providers.

GCP also have GCE, and AWS LightSail.

However, Google Cloud Run has the most simplified workflow. Has to use Docker, though.

[omar shabab]

Certified with both GCP and AWS

GCP UX > AWS UX

At least, I love cloud shell and cloud9 doesn't cut it.

That said, I have a slight preference for AWS because of the community

[Charles Landau]

I can't stress this enough: the differences are not always dispositive!

I am often dealing with someone who has already chosen a cloud, has data in that cloud or an enterprise agreement. Data/cloud "gravity" has set in.

On the merits in isolation?

AWS: market share
Azure: Corp IT
GCP: Predefined workflows (e.g. Firebase family, BQ warehouse)

[Shaiju T]

I have used Azure to deploy web apps.

I have used AWS to create a VM.

• Azure has user friendly UX, its easy to understand.

• I heard that AWS is costly than Azure.

• Recently in Microsoft Build 2020, they announced new AI supercomputer .

• AWS to Azure services comparison is handy to compare as required.

[Alan Barr]

For enterprises I have seen Azure and AWS playing very nicely with all the governance and control required. I've heard that GCP, once you get to a certain size can become unwieldy. There are some differentiators some others have listed in the comments that might be more of a draw for your use case.

[Ido Shamun]

A GCP fanboy right here. Their Kubernetes as a service (GKE) is #1. I think GCP is super under rated if you look at the cost and the services that you get. Also their dashboard is pretty straightforward compared to AWS and Azure where I can never navigate properly.

[e-francis]

Can't agree less

[Vitor Amaral]

You're right the GCP dashboard is the most simple and organized, but Azure is not far away, AWS is a mess

[Rémy 🤖]

AWS — Does everything, super complex because super complete, state of the art

GCP — Slightly simpler, full of Google-specific quirks

Azure — Ow gawd will I ever manage to create an account? How many errors 500 can a mainstream cloud provider get per second? Does anything work in there?

[Rodrigo Asensio]

Maybe as a customer you ended up using a few services and never notice that there is a service by the minute to control satellite comms. And that is OK. Was built for other customers who needed it.

[Gwyneth Peña-Siguenza]

If you're a Windows shop: Azure > AWS > GCP.
If you want to get started learning in order to get a job: AWS > Azure > GCP. There's just more jobs and documentation on AWS than the others, though Azure is catching up quickly.
My personal favorite is Azure, the UI and support has been way better to manage than AWS. GCP UI is overcomplicated.

[e-francis]

I started with GCP,. Now Azure

[Rémy 🤖]

The AWS UI feels like a cold bath of austerity and hardships but in the Azure UI you replace water with liquid nitrogen... Every time I used it I got countless errors on literally everything I did.

[Laurence 'GreenReaper' Parry]

I just finished my first month's trial with Azure, having used AWS for a year before and also using GCP's free instance for Munin monitoring. All three have value, but you have to pursue a multi-cloud strategy and marry it with non-cloud assets for the best results.

If you're looking to be cheap, GCP's free offer excels if you are smart and know how to make the best of what you are given (e.g. you only get 1GB/continent free transfer out per month from your GCP VM, but route that through an app service and it becomes 1GB/day). But you have to work at it.

Azure gives you double the VMs but with the cost that one is running Windows. You can also set up one app service per region, which means 5GB/month per region - some of which are quite pricy for transfer or VMs normally.

All three are too expensive for media serving, which is best handed to dedicated hosting providers such as OVH and LeaseWeb - or if your needs are smaller, various VPS hosts (Vultr, Host1Plus, Civo, Atlantic.net, LayerStack - be wary of SSDNodes, they aggressively trim using the memory bubble to the point of causing increased latency).

Not previously mentioned in this comparison are Oracle Cloud (which has a very generous free offer, if you can actually get the services they are offering and want them in a specific home area); IBM Cloud, which is in various locations that may be useful to you and has DB2 linked to its own app services, as well as object storage; or Alibaba Cloud (which I have not yet tried out myself).