Good Morning!
From the Community
In another exciting episode of "Some PM at Google Runs Their Mouth Unfortunately," they have a blog post up titled, and I am not embellishing here, Cloud Spanner is now half the cost of Amazon DynamoDB, and with strong consistency and single-digit ms latency. Look, you don’t see a lot of Google price cuts and this is great–but that swipe at DynamoDB doesn’t land well. It makes Google feel incredibly socially insecure, it tries to turn a product with significant competitive differentiation into a commodity, it attempts to swap "value" for "price war," gets some statistics about DynamoDB usage hilariously wrong, and to top it all off for many use cases it isn’t even true–last month DynamoDB cost me $2.93 to create this newsletter for you all. Cloud Spanner starts at $65 a month. The only people who feel good about a post like this work at Google; customers don’t find that it resonates, observers cringe, and it harkens back to the olden days of the Database Wars. Do better.
If you missed my post The Cloud Devil You Know, you can remedy that now.
Podcasts
Last Week In AWS: AWS Big Bag of Hammers
Last Week In AWS: Better Late Than Even Later
Last Week In AWS: The Cloud Devil You Know
Screaming in the Cloud: Storytelling Over Feature Dumping with Jeff Geerling
Screaming in the Cloud: When Data is Your Brand and Your Job with Joe Karlsson
Choice Cuts
New Amazon CloudWatch metric monitors EC2 instance reachability to EBS volumes – In yet another mildly confusing turn of events, AWS now lets you know if your EBS volumes are "reachable". I’m left pondering a more fundamental question: When, exactly, did EBS volumes learn to play hide and seek, and why wasn’t I informed? Well, I mean I suppose I was informed when THE ENTIRE INSTANCE SPIKED ITS LOAD AVERAGE TO 700 AND WENT COMPLETELY NON-RESPONSIVE AS EVERYTHING BLOCKED, but I didn’t exactly need a detailed CloudWatch metric to tell me that. Further, I can’t remember the last time an EBS volume went on walkabout. Seriously, what’m I missing here?
Announcing AWS Lambda’s support for Internet Protocol Version 6 (IPv6) for outbound connections in VPC – "Huh, nobody gave a crap about IPv6 support for a bunch of our services, but now they’re screaming at us to give it attention. I wonder why–wait, what? We’re going to start charging for IPv4 addresses in February? Oh, CRAP."
Announcing new AWS Network Load Balancer (NLB) availability and performance capabilities – This unlocks some new capabilities in conjunction with the Route 53 Application Recovery Controller (or whatever it’s named; I’m too lazy to check because I’m on a deadline and it’s too expensive for me when I use it as a database); zonal shift is a game changer for some things, and I’m still shocked that Kubernetes isn’t inherently aware of a similar concept when it comes to AWS AZs.
Two billion downloads of Terraform AWS Provider shows value of IaC for infrastructure management – Nobody is suggesting that there are two billion people who know what AWS is, let alone use it via Terraform. Nobody is suggesting that there are two billion AWS accounts. So what exactly is the point of this vanity metric?
Why purpose-built artificial intelligence chips may be key to your generative AI strategy – If your "generative AI strategy," which you should not have at this time, requires "bespoke silicon only available from a single cloud provider" to effectively execute on, perhaps you should stop letting ChatGPT chair your board meetings.
How Zalando migrated their shopping carts to Amazon DynamoDB from Apache Cassandra – In a typical "everyone look, someone else did it so you should too" propaganda post, AWS is chest-thumping about Zalando shifting from Apache Cassandra to Amazon DynamoDB. Shockingly, they forgot to mention all the late-night hours and half-eaten pizzas that were casualty to this "simple" migration.
Unlocking cost-optimization: Open Raven’s journey with Amazon Aurora I/O-Optimized leads to 60% savings – Y’know what sucks about this? That 60% savings is artificial to AWS’s own billing dimensions, and an outgrowth of Aurora’s IO charge that’s impossible to predict without running the workload to figure out what it’s going to cost you. All Open Raven did here was switch to the I/O Optimized setting–as a result, each Aurora instance costs more per hour, but the I/O charge goes away. This is something that AWS could absolutely provide as a "just how it works" without requiring human involvement, but they don’t. Also, think about what this means: Open Raven was using Aurora, and a MAJORITY of its cost was on the ridiculous I/O charge!
How does Cloud enable the transformation of Bank finance functions? – This feels less like a blog post title and more like a hard question a bank exec asked their AWS sales rep during the initial phone call.
AWS Cloud Institute: Virtual training program for cloud developers – My original take on this program was that it was offensively disgusting, and my reaction was more than a little bit unhinged. Learning new skills? Yes, that’s important, and if that’s where it stopped it’d have my full-throated endorsement. But it costs folks $7500 and a year of their lives to learn what I assumed were entry level skills surrounding a single-vendor’s view of the world, with that money being paid directly to that vendor. That’s a BIG lift for folks fresh out of high school! I reached out to some folks at AWS to gain clarity before shooting my mouth off in public (what a concept!) and found my opinion changing. This is beyond a typical bootcamp approach; there’s a lot of emphasis on business skills, on understanding the interplay of workplace dynamics, and individualized support. Large customers were asking for this. Now, the writeup doesn’t do a great job of explaining the value; in my readthrough I thought the artifacts basically were a Cloud Practitioner certification and some digital badge for LinkedIn or whatever. Instead, there’s a capstone project, a series of discussions with potentially interested employers, and the start of a professional network. So, should you do it? Here’s a quick question to ask yourself if this is appealing to you: imagine you have $7500 and a year to invest into your career. Is this the best use for those resources that you can find? If you’re seriously considering this or anything like it (including a degree, a bootcamp, etc,) before you drop thousands of dollars and that kind of time, find a few people who are doing the job you hope to be doing in three to five years. Take them out for coffee; it’ll be the best investment you can make. Ask them what they wish they’d known when they were in your position, and what they advise you to do. If you can’t find someone who fits the profile? Hit reply and talk to me; I owe a lot of people favors I can only pay forward. Just please don’t hurl your money and more expensively your time at something like this without some very serious consideration. Your future self will thank you.
And my thanks to the folks at AWS who took the time out of their day to walk me through their thinking rather than assuming I was just on one of my nutbar rants again. I like what you’re building; keep going.
Tools
aws-ipv4-cost-viewer shows the future public IPv4 costs for a variety of AWS resources across all AWS regions from an account.
… and that’s what happened Last Week in AWS.
