Good Morning!
Another week has flown by; the kids will be returning to school soon, for which I am extraordinarily grateful.
These are the dog days of summer; enjoy them.
From the Community
A threnody for Amazon QLDB, a service that was amazing but Amazon never quite figured out how to talk about.
Vercel found a cost effective trick to stop paying for Lambda idle time.
The bastards at Broadcom were tidying up after the VMware acquisition and found another rug to pull. In a month the Bitnami container images and AMIs are going away unless you pay them a reported $5K a month. Before assuming this doesn’t apply to you, check your environments. There’s a terrific chance you’re in for an unpleasant surprise, since Bitnami (pronounced "bitten A. M. I.") has packaged a LOT of stuff over the years, done so super well, and become the default for a lot of installers out there.
Podcasts
Last Week In AWS: In the Bleak Theater of the Cloud: A Werner Herzog-Style Dispatch
Choice Cuts
Amazon Aurora MySQL database clusters now support up to 256 TiB of storage volume – It’s become pretty clear over the years that I don’t exactly know what a proper database is or how to use one. If you have a quarter petabyte of data in Aurora, it’s pretty likely that you don’t, either.
Introducing v2 of Powertools for AWS Lambda (Java) – For once Java gets the neat toys before the rest of us do. This is promising!
Introducing Extended Support for Amazon ElastiCache version 4 and version 5 for Redis OSS – So,if you don’t upgrade your cluster, the following happens: for the next two years you’re charged an 80% premium, for year three you’re charged a 160% premium, and at the end of that year you’re… forcibly upgraded to the next Extended Support version where you’ll continue to be charged. All it would take for me to go from "this is AWS grasping for margin" to "this is AWS doing a terrific thing" would be one change: explicitly donate all revenue that this throws off to the upstream projects. I’d become its biggest champion overnight.
Amazon DocumentDB Serverless is now available – No, it’s not; it scales down to 0.5 fake compute units, which means it doesn’t scale to zero, which means it’s not Serverless. How many times do we have to teach you this lesson, old man?
AWS Lambda response streaming now supports 200 MB response payloads – A 200MB response payload is almost enough to hold all of the tracking JavaScript needed for a modern website to display 2KB of text.
How Zapier runs isolated tasks on AWS Lambda and upgrades functions at scale – I don’t want to hear how they run the tasks on them, I want to hear how the hell they keep their email systems from falling over and crashing into the sea under the weight of millions of AWS emails whining about a Lambda runtime they’re using is going to be deprecated in 18 months.
Amazon Application Recovery Controller now supports Region switch – I got a demo of this (because it was too expensive for me to kick the tires on in my test account) recently, and it’s pretty great, provided (and this is a heck of a caveat) your application can handle the shift. This is an architectural rethink, rather than a bolt-on to something pre-existing. It’s wonderful, you should use it, but you’re probably going to have to spend a lot of time making your technical debt repayments.
Announcing general availability of Amazon EC2 G6f instances with fractional GPUs – These start at 20¢ an hour for you to try to spin up, then have your request fail citing insufficient capacity.
Amazon Promotes Malphas to Senior Vice President of Bad Decisions, Unveils 17th Leadership Principle – Probably the most honest press release AWS has ever put out. Of course I wrote it; do you think they’d be this self-aware?
Amazon CloudFront introduces new origin response timeout controls – This is a great enhancement. It made sense back in historical times, but these days? If S3 takes 30 seconds to return a result, it’s not "latent," it’s "dead."
Optimize traffic costs of Amazon MSK consumers on Amazon EKS with rack awareness – Rack Awareness in Kafka is absolutely your friend, because the cross-AZ data transfer gouging is not.
Amazon Bedrock now available in the US West (N. California) Region – This is a great combination of two very different things with the exact same failure mode: neither us-west-1 nor Bedrock has enough capacity for your needs unless you are a Very Special Customer.
New AWS whitepaper: AWS User Guide to Financial Services Regulations and Guidelines in Australia – These days the US version of this isn’t a whitepaper, it’s a whiteboard.
Amazon EC2 Auto Scaling adds AWS Lambda functions as notification targets for lifecycle hooks – Y’know what I’d love to see? A semi-standardized URL format that I could drop into various things on the internet and they just intuitively understood that it was a Lambda endpoint. Think something like the way apprise does things and you’re pretty close.
… and that’s what happened Last Week in AWS.
