Let me point out some cloud magic tonight / say some nice things.

Normally I dunk on @awscloud in these threads, but today that puts me Nazi-adjacent. Plus AWS marketing's nerves are a little... frayed, so I fear for my safety.

The easy starting point is @awssupport. People are just rude as hell to them and they take it on the chin like the professionals they are. "Amazon is owned by a billionaire!" but he doesn't have to field the support tweets.

They also have to figure out when I'm trolling.

No matter how much you complain, they'll help you out. They're marvels, and good people too.

Adjacent to this is AWS's training group. Go check out the AWS product page. It looks like I'm trolling you, right? That horror is *real*!

They teach people to understand all of this--repeatedly. That's no small thing.

The AWS network is likewise magic. Ever notice how in datacenters the top of rack switches like to melt?

Ever notice how you can get full line rate to basically all of your instances all of the time?

Now think of those two things at once and marvel.

Once upon a time, failed hard drives were a "call the datacenter flunky" open ticket item.

Today failed hard drives are an easy-to-miss incredibly brief latency spike. Realize it's been years since you actually cared about SMART.

Behold a series of storage systems, all of which offering you linear pricing on infinite storage.

"Storage isn't infinite!" you might say. It's more infinite than your budget is, I promise.

I complain that us-east-1 has 392 different EC2 instance types/sizes between which you can choose, but that also means that no matter what your workload looks like, there's an instance that'll fit it like a glove.

Some recent things: VPC Reachability Analyzer. S3 Storage Lens. gp3 volumes. All of these are things you should have in your back pocket for when you need them.

Even with their enormous global footprint, there has never been a global network control plane outage for @awscloud in fifteen years. Not once.

The exception case you're about to respond with isn't one of them.

They consider APIs to be promises. Anything you had that worked to provision services in 2009 will still work today. The services are all still there.

I escalate weird customer issues to @awscloud service owners fairly frequently because I neither understand nor respect boundaries.

None of them have ever asked how big the customer was, because it didn't matter.

If you build a site on @awscloud and then leave (set it to autopay first!) for five years, you'll come back to a site whose infrastructure is more reliable, better performing, and less expensive.

Meanwhile the raccoons dismantled your data center.

Have you really stopped to think lately about the idea that you can spin up a world-ranked supercomputer for as long as you need it and no longer, pay for it with a credit card, and then turn it off and be done with it?

That still blows my mind.

People love to complain about service limits, but without them I'd be on the phone with @awscloud three times a year sweating blood that they're going to forgive a typo that cost me $200 million.

I whine and complain an awful lot about @awscloud services. Truth is, 95% of it is awesome, so I don't bother talking about them. It's become a utility; you don't question whether water will come out of the faucet when you turn on the tap.