If I want to learn from someone who really understands how clouds work, who should I listen to?
That's right; @united Airlines. Ethan Rucinski (in association with @awscloud's Ravi Palakodeti) presents "How to reuse patterns when developing infrastructure as code."
(This is a fascinating partnership; AWS *wants* to kill me, and United has the most opportunities to actually do so.)
This is one of those "silent disco" types of talks. It's unnerving to give talks like this. Remember that if you're tempted to dunk on a presenter who's giving one. Honestly, you shouldn't really be dunking on presenters themselves anyway, but I digress.
This is one of those "silent disco" types of talks. It's unnerving to give talks like this. Remember that if you're tempted to dunk on a presenter who's giving one. Honestly, you shouldn't really be dunking on presenters themselves anyway, but I digress.
So far it's just Ravi (AWS Solutions Architect) on stage. If Ethan's not here because his plane was late, he's never going to live it down...
"re:Implement" was going to be the name of an AWS cloud migration conference but it was deemed to be too "on the nose."
Nobody bothered to ask why @awscloud names conferences after email subject lines in the first place.
That slide also talks about multiple Lambdas talking to the same DynamoDB table.
My philosophy has been to stick to a "one datastore per microservice" architectural pattern. I welcome feedback on whether this is sensible or not!
AWS historically never shut up about the CDK and I made fun of them for it. Then I used the @awscloud CDK myself and discovered how awesome it is. Now I never shut up about it either. Well played.
I'd like to hear from United rather than the AWS perspective. @awscloud says a lot of things; I agree with many of them. In most cases they carry a lot more weight coming from the customer instead of the vendor who sells the service.
Okay, now passing to Ethan from United. Thank goodness, this is the part of the talk I'm here for!
WHAT THE HELL AWS. You won't let me beep the horn on the Snowmobile, but United lets their folks hang out with the planes?!
I know where I'd work if I were on the market. 🤬
It sounds like Ethan and I have both flown roughly the same amount on @united.
Whenever United has a problem with the cloud, they can fly up and fix it.
I'm sure nobody has ever made these jokes before.
Note that they're over 120 AWS accounts; this is GOOD. You should be doing this as well.
This is very clearly important. If you want teams to adopt cloud safely, you've gotta make it easy for them to do it. Adding friction means it won't happen.
I'm calling it; the battle of "stop putting -ops after random words to describe things" has been lost.
"We started off by using CloudFormation, and it was great for a while--" okay, the hand of AWS's marketing org spotted. The original word "great" was almost certainly something far closer to "shitty."
Oh my god yes. *SO DAMNED MUCH PAIN* would have been avoided if instead of bespoke crafted names, all actual resource names had been generated by banging on a keyboard. Y'know, how Samsung names monitors.
"Here at @united we're have a technical tradition of inheriting other people's stuff."
Yeah, like Continental's old executive team. HEY-OH!
Their own CDK L2 construct makes getting a DynamoDB table done quickly yet with the right guardrails.
Note the tags are enforced. Developers will NEVER do this manually--at least not consistently. I promise.
Now demonstrating the same thing via the CDK in multiple languages. Every company is inherently a polgyglot shop, so this is important. They publish their stuff automatically to all of their supported languages internally.
Now doing a demo that uses GitHub Actions instead of AWS's CI/CD offerings because the Chicago #awssummit is only one day long.
This hits the classic demo failure mode: these things are designed to be great in daily use, which is very different than "looks good on giant screen demos."
And the @awscloud presenter comes back to put a quote from human garbage fire ESR on the slide, and I'm done. Fortunately so is the talk itself. Hell of an ending note...
• • •
Missing some Tweet in this thread? You can try to
force a refresh
I'm at the AWS Summit in NYC, where I believe that nicknames are for friends--and Gennifer Artificial Intelligence is no friend of mine.
Good morning.
Thirsty much?
A game / challenge at the AWS Startups booth: how long can an AWS employee go without mentioning GenAI? Someone just made it all the way to one minute, ten seconds!
Okay. Let's do Networking Specialty. Practice question 1:
Correct answer is B.
"Wrong!" says the answer key, "it's B because network load balancers don't support client IP preservation."
Except that they do. They absolutely do. They have for the past year. I'm just a boy, standing in front of an AWS Cert team, asking them to do their damn jobs.
Today's cloud marketing story is called "The Tale of Hot Rebecca," and is a truthful recounting of dinner last night.
Strap in; it's a fun ride.
Back in my early 20s, I had a number of friends / acquaintances in my (primarily Jewish) social circle named "Rebecca." It was kind of a problem.
("Can't we spray for them?"
"…not since the 1940s.")
So every Rebecca got an adjective, much like the seven dwarves. One of them asked me once what her adjective was, and I responded in a fit of unadulterated honesty, "you're Hot Rebecca" because honestly? Damn.
Made it to the #GoogleCloudNext keynote seating finally. Let's see how this goes now that the world is starting to wake up to a "much of the AI hype is unwarranted" reality.
Boeing: "HOW ARE THEY DOING IT?!"
Airbus: "We bought a torque wrench?"
Boeing: "No, how are you being a featured customer testimonial at #GoogleCloudNext?"
Airbus: "Oh, that? We made a strategic decision to not be walking poster children for corporate negligence."
And now, some DevOps / SRE / Sysadmin / Ops / ENOUGH already tips I learned from early in my career--brought to us by our friends at Chex™ Mix. All of these are great ideas that you should implement immediately...
DNS is notoriously unreliable, so use configuration management to sync all of the servers' /etc/hosts files. Boom, no more single point of failure.
Future-proofing is an early optimization, so don't do it. Every network should be a /24 because that's how developers think. I mean come on, what are the odds you'll ever have more than 253 hosts in a network?
And the Amazon earnings are out for Q4. A miss on @awscloud revenue by $20 million because analysts didn't expect one of you to turn off a single Managed NAT Gateway.
Let's explore deeper into their press release.
For 2023, AWS sold $90.8 billion of services, most of which were oversized EC2 instances because you all refuse to believe Compute Optimizer when it tells you there are savings to be had if you're just a smidgen more reasonable.
Word frequency in the earnings release:
Customer: 87
Employee: 11
Generative: 16
Cloud: 24
Serverless: 3
DynamoDB: 2
Union: 0