Good Morning!
I spent part of last week at Google Cloud Next, which was a great show if you’re super into AI boosterism but a terrible show if you instead care more about ancient things like "Kubernetes" or "running cloud infrastructure in production."
If this is a bellwether, 2024 might be an excellent year to skip re:Invent.
From the Community
I’m hosting a podcast recording before a live studio audience at RSA next month. If you’re around San Francisco, I recommend signing up for this one. I’ve always wanted to mouth off in front of people in person!
Charles Fitzgerald is once again on fire as he tracks the Clown Car Race Checkered Flag of cloud CAPEX.
Microsoft employees exposed internal passwords because of course they did. There’s something deeply and profoundly wrong with that company’s entire security culture.
This thoughtful piece on Any Technology Indistinguishable From Magic is Hiding Something is one of the best analyses of the current state of AI in the cloud that I’ve come across. It explains a lot.
I did not enjoy the first version of this post: We discovered an AWS access vulnerability. The original timeline (since updated!) suggested AWS took two weeks to acknowledge the initial reporting email, which is "thank you for calling the Azure Security Desk" platinum tier response time, but glacially slow by AWS standards. I’m glad the timeline got updated: they acknowledged receipt the next day. That aside, it’s a great find and a complicated issue; I remember predicting that suddenly pivoting to using tags for security would cause pain when it first shipped…
Y’know, I’ve often said that I don’t know of too many companies who spend more on cloud services than they do their employees, but Stability AI reportedly ran out of cash to pay its AWS bills • The Register because they’re apparently the exception case. This isn’t how the AI bezzle works! FIRST you milk the VC firms like desperate cattle, THEN you light the money on fire for GPUs. If you do it in the other order, you run out of money and your business burns to the ground. Worst of all (according to the people on this newsletter who work at Amazon), the final AWS bill might–gasp–go unpaid!
Wow, I didn’t expect to get formal confirmation from AWS that datacenter power issues were causing capacity constraints in their Dublin region, but the complete non-answer from their PR person may as well have been a "yup, that’s what’s going on." Seriously, go read the article and the official PR statement; it demonstrates super well why tech firm PR has become a running joke.
Man, "us-east-1 is a tire fire" sure is a dated reference. It used to be, yes! Then it got fixed, and AWS put out an awesome whitepaper about fault isolation / things that single track through the region, and now it seems like this article got stuck in the editorial queue for five years.
Podcasts
Last Week In AWS: Get Billed For a G6
Screaming in the Cloud: Crafting Tech Success from Bad Ideas with Xe Iaso
Choice Cuts
Amazon Route 53 adds support for 18 additional Top-Level Domains – Your database has expanded its options for valid namespaces yet again.
Announcing AWS Transfer Family workshop for building secure file transfer solutions using SFTP – Perhaps I’m old, but setting up an FTP server was never the sort of thing that required "a workshop" when the world was young. Just how overcomplicated have they made the Transfer Family since I last looked at it, anyway?
New AWS monetization solution demonstrations at NAB Show 2024 – Ah, the Amazonian shitbirds responsible for slapping ads on everything that even slightly holds still long enough have expanded beyond their last target of "the AWS Marketplace" (seriously, it has sponsored results now, which is absurd) and decided to evangelize this to other companies now as well. I’m increasingly of the mind that advertising is a corrosive force to big tech companies, as it invariably ruins what were previously not horrible experiences.
Optimizing AWS Backup costs – I hoped for more from this article; unfortunately it’s "understand what things cost and along what dimensions," "make sure you’re backing up the stuff you need to back up," and "ensure you’re not keeping every backup forever, because that gets really expensive." This is… not exactly revelatory.
Migrating More than 250 Billion Daily Connections to AWS Network Firewall with AWS | AWS Case Study | AWS – AWS had a boastful post about moving their networking to Managed NAT Gateways and its more expensive cousin, the Network Firewall. At an absolute floor this thing costs over $200 million a month just in data processing fees at the prices you or I would pay. Why would they do this? “Expanding capacity in hardware firewall projects often took 6–12 months. With rigid, expensive hardware firewalls that require specialized equipment, migrating and scaling involves additional steps that don’t exist when using AWS Network Firewall,” says Wade Millican, senior manager at AWS, whose team is responsible for operating the company’s internal firewalls. Or so we think; it was hard to hear him clearly as he was backing a dump truck loaded with gold bricks into the NAT Gateway product manager’s driveway.
… and that’s what happened Last Week in AWS.