Good Morning!
Another week has come and gone; AWS is getting back into its groove, and so am I…
From the Community
This is sad: AWS to Shut down Aurora Serverless v1. That’s the version that actually scales to zero; not scaling something to zero is neither serverless, nor is is welcoming to a whole bunch of use cases that customers have grown to love.
The Information has a (paywalled) article about AWS doing yet another sales reorg. What’s fascinating is that they’ve gotten numbers: AWS has 115K people, 60K of which are in their sales/marketing org. Of course, AWS says those numbers "are not accurate," but unless they can provide correct figures then I counter with "okay, sure, those numbers may be off by a dozen people in either direction."
Podcasts
Last Week In AWS: NewYear, New You, Here’s December in Review
Screaming in the Cloud: Championing CDK While Accepting the Limits of AWS with Matthew Bonig
Screaming in the Cloud: The Importance of the Platform-As-a-Product Mentality with Evelyn Osman
Choice Cuts
AWS Accounts discontinues the use of security challenge questions – "Why did these get discontinued?" "Nobody asked." This always felt like security theater; I’m glad to see it being removed.
AWS CodeBuild now supports a X-Large Linux compute type – This remains the best actually-serverless offering that AWS has to run containerized jobs on a schedule. This lets you do more with fewer jobs, and I’m not being even slightly sarcastic here. There are a universe of monolithic build pipelines that are configured in "interesting" ways; this meets customers where they are. CodeBuild has long been an underappreciated AWS service. I’m glad it’s getting some love.
Introducing Open Job Description, an open specification for portable render jobs – If I told you that I was launching something called the "Open Job Description," you would think I was talking about hiring humans. Nope! AWS is using this to talk about "Render jobs," which you’d really think they’d have worked into that title to avoid mass confusion.
Reducing Inference Times by 87% for Darwinbox’s Talent Search Engine Using AWS Inferentia – Hey, listen, since AWS is obviously singing Inferentia’s praises every time it does something, why isn’t Amazon Q in that list? Are they holding back NVIDIA GPUs that customers are clamoring for, to advantage their own inept chatbot? Also, given that basically nobody is using Inferentia at scale, what was Darwinbox doing before, inference on CPUs? An 87% increase is… suspicious, particularly when they don’t talk about what they were doing before.
Amazon ECS supports a native integration with Amazon EBS volumes for data-intensive workloads – I was surprised last year to learn that Fargate didn’t support EBS volumes. Congrats to AWS on doing the thing we’d all assumed it had been doing for the more than six years since it launched?
Tools
The BBC has released an SQS Consumer that makes working with this venerable service a lot less unpleasant / full of boilerplate.
cdk-goat is a "vulnerable by design" infrastructure that can be deployed for training purposes–or presumably if you work at Okta, for production use.
Using S3 as a database with SQlite is delightfully cursed.
… and that’s what happened Last Week in AWS.
