Thread by @QuinnyPig

I've been hoping someone would build something like this to show relative @awscloud service costs. A thread!

https://twitter.com/JoinVantage/status/1417529953181184002

"Why not you?"

I can't very well do it myself. At The @DuckbillGroup, our clients these days start at ~$1 million a month in spend or so. I'm very hesitant to give guidance to small accounts based upon what large ones are doing. It's a very slanted view of the industry!

That said, the data I'm seeing in here tracks with what I'm seeing in our client environments. As the post says, "this aligns with other cloud consulting organizations @getvantage has spoken to." We're one of them. They're spot on for the big items.

I'd like to draw attention to the "EC2-Other" category. That covers data transfer and EBS (each of which would be top five in my own analysis). And of course, Managed NAT Gateways.

And snapshots and IPs and the tea lady's tips and...

We're also seeing significant shifting from "EC2 instances" to other forms of compute--specifically Lambda and Fargate. It's challenging to determine how much of that is net new vs. EC2 workloads migrating over time.

This also provides a glimpse into my world. See how EKS is down at number 20? That's purely Fargate + the (trivial) cost of the EKS control plane. EC2 worker nodes remain EC2.

Same story with ECS.

I don't usually pay attention to the very long tail (it's ~20% or less of spend), but this is fascinating to theorize about.

Route 53 and ACM are both bigger spenders than SageMaker.

It also gives a view that's unhelpful. "Athena is all the way down at #61, how serious could it really be?"

The data it's querying all lives in S3, jackhole. There's an idea of "to use service X, you pay a boatload to service Y."