Thread by @QuinnyPig

Ooh, the @awscloud Application Cost Profiler has come out, and I'm going to test it.

aws.amazon.com/about-aws/what…

It wants an S3 bucket. Cool. The service has one bucket for ingest, and one for output generally. Which does this form want?

ALWAYS WITH THE QUESTIONS, YOU PEOPLE!

"Encryption of S3 buckets is basically a box check for compliance groups. That said, we at @awscloud are going to mandate it for this service with your own KMS key because the KMS team bribed us for that sweet $1 a month revenue juicer."

I have created a $12 a year @awscloud KMS key that this product requires me to have, and named it "goan" as in "goan fuck yourself!"

And I have to manually edit the key policy as well.

Now that that's done I have to--wait are you being serious right now?

"Go to the console and configure your report."

The only thing in the console is the screenshot earlier in this thread. What on earth...?

I'm having a hard time believing what I'm reading. "Go ahead and configure the table that gets uploaded to this service. Each application has to report to this thing how much of each resource it's using.

You basically have to build this thing yourself.

Then it charges you $5 per GB that it ingests and "processes," which looks a lot like "basic arithmetic."

This entire process should be replaced with a button labeled "smee" in the console.

When you click it, you get a popup saying "smee again, goan fuck yourself!"

Why is this any better / easier than just pushing your own metrics to Athena? Or tagging resources and then using Cost Explorer (or Tableau, or whatever) to slice and dice the CUR yourself?

If I have to build the thing from scratch, wtf is the service *doing* for me?

I have deleted the bucket & scheduled the KMS key for deletion, but there is no way in the console to delete the Application Cost Profiler configuration itself.

I do not understand this service, its market, its use case, or its service team.

Strong avoid.

It's not, though! Who the hell is going to configure this thing, then ensure that all the applications write to this dingus instead of, y'know, existing systems?

https://twitter.com/MarioHachemer/status/1395116832307171333?s=20

And you've gotta instrument your applications to write to this thing!

If you're already doing *that*, just have it emit a metric to some database somewhere with its application ID, then roughly apportion shared resources based upon hit count / timestamp?

It'd be less work.

I've carefully reviewed the logos at duckbillgroup.com/clients and have failed to identify a single company that would not hit me with a belt if I proposed adopting this thing. It's that bad.