This Thursday I’m speaking at an FTC Tech Summit along with other, more qualified people. This should be fun; tune in!
From the Community
AWS SVP James Hamilton has an amazing slide deck from a talk he gave last week. I always learn so much from his talks; a gem from this one is that EC2 P5 instances cost AWS roughly a third of a million dollars, and take up 6 rack units.
This post on Slashing Data Transfer Costs in AWS by 99% distills down to using S3 as an end-run around cross-AZ data transfer charges. It’s a good pattern if your use case supports it; it’s also ridiculous that you have to play these sorts of games.
Last Week In AWS: Deprecations and a long-awaited Fargate feature
Screaming in the Cloud: Continuing to Market After the Product Has Sold with Kim Harrison
Screaming in the Cloud: The Future of Entertaining Developer Content with Jason Lengstorf
AWS Transfer Family provides static IP addresses for SFTP connectors – Historically you had to do hacky things with NLBs and whatnot to make this happen. Great enhancement, but years after a bunch of folks really needed it.
GLIDE for Redis, an OSS Redis client sponsored by AWS, now available in preview – I can’t for the life of me figure out if this is due to a licensing thing, or because the open source Redis client landscape is a wasteland. Maybe both? Either way, it’s a new day.
AWS Marketplace features to accelerate your renewals strategy – Future Dated Agreements are the real stars here; you can now plan ahead. Of course, planning ahead is anathema to many companies, but at least it’s harder to blame AWS for it now.
Amazon EKS extended support for Kubernetes versions pricing – I love this so much. Now, 14 months after a version of Kubernetes gains support in EKS, it shifts from costing 10¢ per hour per cluster to 60¢. This is GREAT. Kubernetes is clearly a fast-moving, rapidly evolving technology; if you implement it, you should have a plan for how you’re going to keep your infrastructure current. If you take a lazy approach to it, you’re causing negative externalities for the rest of the internet as your lapses start to take on the shape of "inadvertently providing infrastructure support to bad actors." This helps shift some of that cost onto you if you’re doing it. More like this, please.
Save up to 90% using EC2 Spot, even for long-running HPC jobs – This article starts with the phrase "But long running HPC jobs often can’t survive an EC2 Spot interruption which makes it hard to benefit from all this low cost and large volume of compute" in its first paragraph. Trouble is, the solution they talk about doesn’t… actually fix this? AWS needs Spot to make HPC even slightly cost competitive with on-premises data centers, but so many HPC workloads flat out can’t tolerate instance interruptions.
OpenSearch Expands Leadership Beyond AWS – In a move that shockingly acknowledges that open source software might involve people beyond Amazon, OpenSearch has decided to expand its leadership team outside AWS confines. Just in case we questioned their commitment to ‘open,’ they’ve gone for the bold strategy of actually including other people–but not too many of them. Amazon/AWS account for 7 slots, versus other companies totaling 6 members. It’s still an Amazon project, and it’ll be perceived that way for some time to come, I suspect.
The BBC has put out this helpful SQS Consumer to remove an awful lot of the boilerplate behind building SQS-powered apps.
… and that’s what happened Last Week in AWS.