Good Morning!
After reading it for over 25 years, I finally wrote a column for The Register, obviously about the us-east-1 outage we started last week off with. More to come on that particular front in the near future…
From the Community
Someone finally did it: Datadog launched updog to show whether there’s a massive internet event currently ongoing so we don’t have to trust the cloud status pages to update in a timely manner.
The tech industry’s dirty secret isn’t AI itself—it’s that nobody in a position of power dares admit it’s just another tool, not a religion. When your CEO is placing billion-dollar bets on hype while threatening your job for skepticism, suddenly everyone becomes a true believer. Fear makes excellent marketing. I’m glad Anil Dash called it out so eloquently.
Podcasts
Last Week In AWS: Catching Up, Cashing In
Choice Cuts
Streamline in-place application upgrades with Amazon VPC Lattice – VPC Lattice solves real migration headaches, but calling it "streamlined" when you still need weighted target groups, listener rules, and Gateway API controllers is generous. At least it beats manually juggling ALBs and Route 53 during a botched EKS upgrade at 3 AM.
Build a proactive AI cost management system for Amazon Bedrock – Part 2 – This is a two part article on how to build a system to track expenses faster than the AWS billing time out of Lambda functions, CloudWatch metrics, and popsicle sticks. It takes less time than reading the post to get that outcome with Anthropic’s API natively. Bedrock is really coming from behind, being hampered by the very AWS systems that empower it.
Overview and best practices of multithreaded replication in Amazon RDS for MySQL, Amazon RDS for MariaDB, and Amazon Aurora MySQL – Nothing quite says "we’ve been running databases for 15 years" like a blog post explaining how to make replication not terrible. These are settings you should’ve had optimized by default, not tucked behind parameters that require a PhD to configure properly.
AWS announces Nitro Enclaves are now available in all AWS Regions – Finally rolling out everywhere after being "generally available" since 2020. At least there’s no nickel-and-diming here, since you just pay for the EC2 instance. A relic of a simpler time! That said, convincing your apps team they need isolated compute environments is the real challenge, not regional availability.
Amazon CloudWatch Synthetics now supports bundled multi-check canaries – Stuffing ten monitoring checks into a single canary is some horrifying taxidermy.
Amazon U7i instances now available in Europe (London) Region – Six terabytes of RAM in London because apparently someone’s SAP HANA instance needed its own postcode. At roughly $73 an hour, this is "datacenter in a box" level expensive.
Amazon Connect now supports automated follow-up evaluations triggered by initial evaluation results – Recursive evaluations mean your contact center agents now get scored on getting scored. At least the automation ensures consistent micromanagement at scale, which I’m sure will do wonders for morale and definitely won’t create perverse incentives around gaming the metrics.
How the Wildlife Conservation Society uses AWS to accelerate coral reef monitoring worldwide – AWS built a genuinely useful coral reef monitoring system that doesn’t charge marine biologists by the API call. The real story is that that nonprofits can actually afford to run meaningful ML workloads without selling a kidney, provided AWS foots the bill. Good on them; this cause is important.
Amazon MQ is now available in AWS Asia Pacific (New Zealand) Region – More excitingly, DynamoDB is now available in us-east-1.
Amazon CloudWatch introduces interactive incident reporting – Specifically, they did so the day after a massive incident took them down for half a day in Virginia, apparently without a trace of irony.
AWS Secret-West Region is now available – "What does this cost?" "That’s classified" is some genius level gouging. It always feels super weird to talk about something labeled "Secret" in public, but when has naming things well ever been an Amazonian strength?
Charting the life of an Amazon CloudFront request – This is a technical deep-dive blog post, not a service announcement. It’s AWS explaining how CloudFront actually works under the hood, and it’s super interesting.
Tools
Steve Klabnik leaving Oxide for a jj startup feels like 2012 all over again: betting early on infrastructure tooling that makes engineers’ lives measurably better. He’s notable in the Rust world, so of course the post spends the first third talking about Rust in accordance with that community’s tenets.
… and that’s what happened Last Week in AWS.
