Last Week in AWS Logo

Good morning!

Welcome to issue number 73 of Last Week in AWS.

I'm off to Germany today to give a keynote at SREcon EMEA. The conference is sold out, but I'll be spending the week in Düsseldorf with a few evenings free if anyone is up to grab a coffee, beer, or snack.

Let's see what AWS has in store for us--but first...

Continuous Delivery is obviously a good thing, and if you're not doing it, your company should remedy that immediately--or so "common wisdom" goes. You can't improve what you don't measure-- and this week, Last Week in AWS sponsor GoCD weighs in with 4 Important Metrics for Continuous Delivery. Measure twice, cut once, as the saying goes...

Community Contributions

Bernard Golden (CapitalOne's VP of Cloud Strategy) has written a fascinating dive into the vision behind the beauty that is the Snowball Edge with EC2.

I wrote a brief post about the t3 instance launching-- specifically, a notable pricing caveat.

A nuanced discussion of Serverless vs. Containers is a clever technique to keep developers arguing about pointless questions rather that getting actual work done. Well played...

A quick-and-dirty guide to pulling off a deployment of Jenkins to Elastic Beanstalk.

A detailed price analysis of AWS Lambda and Lambda@Edge. One of these days I should really try my own hand at a pricing analysis...

I missed this when it first came out, but Teads talks about their real-life AWS cost optimization strategy. I'd like you to look at this for a minute, and just think about what other things these engineers could have worked on instead of jumping through hoops to solve what amounts to a global problem, locally. That's my single biggest indictment of AWS's billing model. It's not the cost, it's the complexity, and the constant feeling that you're missing something.

AWS pitches in for research about Amazon deforestation. To be clear, we're talking about the rainforest problem, not a product that was released that flew underneath the radar called "Amazon Deforestation."

This week's S3 Bucket Negligence Award goes to one of the worst kinds of companies-- a spyware manufacturer. They exposed terabytes of selfies, text messages, and location data, none of which the exposed victims wanted them to have in the first place. How many more of these will we see before something changes?

Given the sheer number of people needing medical attention after receiving their AWS bill, it only makes sense that Amazon hired a star cardiologist.

A thorough dive into AWS Parameter Store, one of the hidden gems of AWS. It's useful, it's a best practice, it's free to use.

Using AWS EC2 instance store vs EBS for MySQL: how to increase performance and decrease cost - Percona Database Performance Blog - Percona has a blog post that talks about getting better performance for databases out of EC2 instances than from RDS. I'm not sure that's the right direction to go in (particularly considering there's no Percona RDS option I have to consider the source), but it's interesting metrics that may apply to a number of database workloads. You decide...

Trek10's blog features a deep dive into the nuances of Fargate pricing in context. This analysis "gets it" beyond the pure infrastructure costs.

Do you have a technical story to tell? Would you like to get paid to teach other people how you did it? Would you like a matching donation for what you're paid to go to a worthy non-profit? Consider writing for DigitalOcean's Write for DOnations (no, I didn't typo that--it's "clever") program. You draft an article on how to achieve a technical outcome, you get paid, the non-profit gets paid. I've used a number of their tutorials to implement things myself-- it's a very useful resource. Thanks again to DigitalOcean for their support of this newsletter.

Choice Cuts From the AWS Blog

The Amazon Chime Web Application Can Now Dial Out to Your Phone for Meetings - "Hi, it's Amazon Chime calling! At the tone, please prepare for a chorus of 'who just joined?'"

Amazon EKS Supports GPU-Enabled EC2 Instances - "Unfortunately, the GPUs aren't going to be made available to the containers running on those instances for the foreseeable future" isn't true, but I half expected to find it anyway.

Amazon ElastiCache for Redis adds support for in-place version upgrades for Redis Cluster - You can now upgrade your ElastiCache cluster for Redis in place. Unfortunately you can't upgrade them to something that isn't Redis at this time...

Amazon Lightsail Announces 50% Price Drop and Two New Instance Sizes - LightSail is now dirt cheap, and offering more options. I suspect this service exists solely to taunt us with how straightforward an AWS offering could be-- but isn't.

Announcing New Custom Analysis Features for AWS IoT Analytics with Custom Container Execution for Continuous Analysis - IoT Analytics, now with containers. This is why we can't have nice things.

AWS Key Management Service Increases API Requests Per Second Limits - At 10K requests per second, times the KMS request charge... carry the two... okay, under the default limits, you can now incur charges of $78,840 per month, per primary region. Plus a dollar per customer managed key, but really--that's not the tipping point here.

Introducing Amazon EC2 T3 Instances - I'm still dissatisfied. With this new T3 family, there are now only 140 different instance family / size type combinations I can choose between. There's still too much chance that I'll be able to figure out what works best for my use case. More options, please!

Use Amazon DynamoDB Local More Easily with the New Docker Image - DynamoDB local is a java implementation of DynamoDB for local development (read as: you'll stop 20 minutes in if you knew what it would cost to run this in the cloud). You can now shove it into a Docker image and abstract even more complexity away. Rejoice!

Managing your expenses with Amazon Lex - A do-it-yourself approach to handling expense report transcription and recording, because the existing applications in this space are either expensive or problematic.

Scalyr has published an agnostic (by which I mean "isn't a thinly disguised sales pitch") guide to evaluating log management systems. Be forewarned-- the more I read this, the angrier I become with CloudWatch Logs and its shortcomings. Thanks again to Scalyr for supporting this newsletter.


This kube resource report displays k8s cluster and pod resource usage, runs some numbers, and spits out a per-application cost. I really like the direction this is going in.

Syncing JIRA tickets with AWS support tickets sounds like something you'd have to do on the third level of hell, but here's a way to do it if you're unfortunate enough to need it.

…and that’s what happened Last Week in AWS.

I’m Corey Quinn. I help people significantly reduce and understand their AWS bills and speak broadly on the conference circuit. I advise companies doing interesting things in the cloud space, such as ReactiveOps.

If you’ve enjoyed reading this, tell your friends to sign up at (or post a link in your company Slack team!) about it. As always, if you’ve seen a blog post, a tool, or anything else AWS related that you think the rest of the community should hear about, send them my way. You can either hit reply– or join the #lastweekinaws channel on the og-aws Slack team.

List archives are always available at