Good Morning!
I’m in Las Vegas this week for Google Cloud Next; come say hi tomorrow (Tuesday) night at Atomic Liquors, starting at 7PM. Let me buy you a drink while you confess your secret technological woes to me.
From the Community
Oof; layoffs are never fun, and it would seem that AWS’s Training and Certification group just lost a few hundred people. Those folks have an absolutely impossible job, and they’re remarkably good at it. You’d be lucky to snap them up to work with you.
My gut reaction to the news that Reddit Migrates Media Metadata from S3 and Other Systems into AWS Aurora Postgres is that there’s clearly context I’m missing, because as a general rule the economics of S3 tend to significantly beat almost every other architecture. That said! Reddit does not hire fools, and reserve "lighting money on fire" to non-technical arenas, so there’s almost certainly context that didn’t make it into this write-up.
My Terrible Ideas in Kubernetes talk at SCaLE has been posted on the YouTubes.
Podcasts
Last Week In AWS: Redis Forks and Retroactive Cost Tagging
Screaming in the Cloud: Shifting from Observability 1.0 to 2.0 with Charity Majors
Screaming in the Cloud: The Intersection of AI, Security, and Cloud with Alyssa Miller
Choice Cuts
Amazon GuardDuty EC2 Runtime Monitoring is now generally available – Well this feels like a giant step backward: this needs an agent to be installed on the instance, and it charges a very 1990s-era $1.50 per vCPU. If I wanted to do per-core licensing math I’d probably be an Oracle customer.
Amazon SageMaker Canvas announces new pricing for training tabular models – Oh for… the announcement states "For example: A quick build model with 16 MB of data with about 3 million cells can now be as low as under $2, and a standard build model, which runs a comprehensive AutoML experiment may use 2 ml.c5.18xlarge instance hours and still cost you only $7.30." YOU FORGOT THE $1.90 AN HOUR SESSION CHARGE FOR THOSE TWO HOURS! And you people still owe me my $260!
Introducing AWS CodeConnections, formerly known as AWS CodeStar Connections – A rare glimpse into something AWS has little practice at while Google leads the world: deprecating a service people depend upon. All of CodeStar will be deprecated at the end of July.
Amazon CloudWatch now supports cross-account anomaly detection – Honestly, something in CloudWatch easily working cross-account is the real anomaly. More like this, please.
Amazon EKS extended support for Kubernetes versions now generally available – This is less a feature, and more of a billing update: keep current or pay 6x per cluster control plane. This is on balance good for the internet, but it’s gonna really catch some customers by complete surprise. Yes Virginia, you really do have to update your infrastructure at least once a year.
Announcing AWS Deadline Cloud – What’s interesting to me about this launch is that it goes very well with Amazon’s Nimble Studio offering, a managed service for visual content creation–but none of the blog posts talking about Deadline Cloud mention it at all. What’m I missing?
AWS Console Mobile Application adds support for CloudWatch custom dashboards – I’d give a lot for there to be a reasonable way to log into the mobile app without having to go through the SSO back and forth dance every time I want to check something. The only option I can see is permanent IAM keys, but that’s scary stuff…
AWS Lambda adds support for Ruby 3.3 – I was under the impression that Ruby users were left to largely fend for themselves in "put it in a container yourself" land, but nope–the runtime has been updated.
Announcing general availability of Amazon EC2 G6 instances – ♪ ♫ ♬ Runnin’ workloads on AI, like a wizard ♪ ♫ ♬ Pay NVIDIA for GPUs, through your gizzard ♫♫ (Only get ’em in three regions out in the sticks) ♬ Now you’re gettin’ billed for a G6 For a G6 ♪ ♫ ♬
Announcing per-second billing for EC2 Red Hat Enterprise Linux (RHEL)-based instances – EC2 rolled out per-second billing effective October 7, 2017. "Six and a half years behind current" feels about right for Red Hat post-IBM.
Why AWS Supports Valkey – AWS can’t really win; they’re in no way sympathetic and will have their motives intensely questioned, but they’re doing all the right things here.
AWS Activate credits now accepted for third-party models on Amazon Bedrock – AWS: We’re happy to announce that your Activate credits can now be wasted on third-party models on Amazon Bedrock! Because what startups really need is a convoluted way to burn free credits on marginally useful pre-baked machine learning models. Strap in and enjoy your journey of over-complicating simple processes! <– this entire paragraph was written by GPT4, which isn’t in Bedrock.
As a human, my observation is a lot more snappy: Bedrock’s third-party models can be used for free because they suck.
DoorDash saves millions annually using Amazon S3 Storage Lens – I reiterate that S3 Storage Lens is awesome; unfortunately that’s largely because without it observability into your S3 usage is abysmal.
… and that’s what happened Last Week in AWS.
