Good Morning!
Happy Tuesday! I’m keynoting NANOG in Kansas City in a couple of weeks should you be around and want to hear me opine on networking, in "TCP Terminates on the Floor: The Ebbing Tide of Networking Expertise".
From the Community
Sluggish growth and blindsided by generative AI: Adam Selipsky’s predictable departure from AWS says ITPro, and I don’t think I buy it. In hindsight, most things are predictable–but I was legitimately astounded to hear of Adam Selipsky’s departure. While I like Matt Garman very much and would (and did!) champion him as the CEO of AWS, I didn’t expect it to happen on such a short timeline. I also think that it’s unfair to AWS to fault them for not predicting GenAI’s rise. I do think it’s fair to mock them for their fumbling attempts at rewriting history around it, but they genuinely do not need to be the leader in every aspect of modern technology. It’s okay to miss a trend. Breathe; it’ll be okay. If that’s truly the reason Selipsky was shown the door, I’m disappointed in Amazon’s leadership for the decision.
The Register reports that Starlink offers ‘unusually hostile environment’ to TCP, which tracks. By all accounts, the company also offers an ‘unusually hostile environment’ to employees, which is generally what happens when the CEO is a raging douchebag.
Jobs
Duolingo is on the lookout for a top-notch Platform Engineer to join their team in Pittsburgh. Relocation to the Steel City is required, but it’s known for being a cool and affordable town. If coding, scaling systems, and keeping platforms running seamlessly is your passion, check it out.
Podcasts
Last Week In AWS: AWS Upgrades its CEO
Screaming in the Cloud: Generative AI, Tech Innovations, & Evolving Perspectives with Randall Hunt
Screaming in the Cloud: Teaching a Stanford Cloud Course with Aditya Saligrama
Choice Cuts
Introducing Amazon EC2 C7i-flex instances – Ah, AWS just wheeled out the EC2 c7i flex instances, because clearly, the world was anxiously awaiting ‘yet another instance type’. Now we can all burn up countless hours debating over whether this new shiny toy fits our workload better than the 275 other instance options! I legitimately don’t know where exactly this fits in yet, because AWS Marketing appears "willing to be misunderstood for long periods of time."
Amazon RDS for Db2 introduces hourly licensing from IBM through AWS Marketplace – I tried to get access to Db2 in RDS when it launched. To do so, I had to reach out to IBM Support through a web form. I’m still waiting to hear back six months later. Or, as IBM would likely put it, "why am I in such a rush?"
Amazon VPC Lattice now supports TLS Passthrough – This is a big deal, TCP can now remain yours instead of entrusted to AWS entirely. That said, on Lattice: I mentioned Lattice last week and a couple people emailed to tell me that they want to like it, it’s just too expensive. They made the same mistake I did–assumed that the charge was ON TOP of the data transfer charges between AZs rather than a replacement for them. Could someone on the team please make this more prominently clear so I don’t have to be your unpaid customer ombudsperson on this point?
AWS CloudFormation streamlines deployment troubleshooting with AWS CloudTrail integration – Wow, a link to where the issue happened shows up in the CloudTrail logs. There’s no way that this should feel as transformative as it does, yet we’ve collectively failed at this as an industry for decades.
AWS CloudFormation accelerates dev-test cycle with a new parameter for DeleteStack API – Wrong behavior. The example it gives and that the rest of us have lived is when a stack fails to delete because of a non-empty S3 bucket, leaving the stack in a delete-failed state. The CORRECT behavior is to provide an override that says "No, I get it, delete the bucket anyway because I know what I’m doing–or at least, believe that I do." What this instead does is deletes the stack and orphans the bucket. Bad! Bad AWS! Go and clean up the mess you made on the floor.
AWS Lambda console now supports sharing test events between developers in additional regions – Oh my god do I hate the entire Lambda test event perspective. Yes, I get it, sometimes there are complex payloads you need to provide, but here in my world? I use Lambdas as modified cron jobs. It sorely needs a "just invoke the damned thing with an empty payload without making me jump through hoops" button.
Mail Manager – Amazon SES introduces new email routing and archiving features – This feels like an enterprisey (read as: expensive) version of CloudFlare’s email routing. I have many questions ("does it accept-then-bounce email, thus creating a problem for the rest of the internet") and few answers as of yet.
Join us at the AWS World IPv6 Day Celebration – This is great–an in person event to learn about IPv6. I’m going to attend the one in SF and ask pointed questions about why the AWS Glue team has apparently never attended one of these before.
5 ways to increase AWS Certified employees in your organization – "Did you know that 95% of customers consider cloud certifications important in their decision-making process when choosing an AWS Partner?" I didn’t realize 95% of organizations were delegating their decision making to fools. Don’t get me wrong, certifications serve a purpose, but the idea that everyone cares about them is patently false. Now let’s be clear here–to be an AWS partner of any standing you must have certified staff, so yeah, I could see that 95% of customers don’t want a partner that lies about their partner status, but that’s not the implication of the statement.
Tools
Ooh, PostgreSQL (pronounced, as always, "Post GRUH Squeal") is getting an open source time series extension. This aligns with my world-view: I want one database to rule them all, that can handle multiple modalities. I don’t want to have to pick between two dozen managed database options for each workload.
… and that’s what happened Last Week in AWS.