Our serverless superhero this week is Tanya Boiteau, principal product manager of Step Functions at AWS. Tanya is one of the primary drivers behind the innovative features of the powerful AWS service we know and love, enabling so many critical workloads around the world. Thank you Tanya, for your hard work, dedication, and wonderful ideas!
We’re going to start with a post everyone needs to read if they’re interested in using Amazon Bedrock. Matteo Depascale has put in some hard work and detailed how to use every model currently available in Bedrock. Something I learned recently is that the inputs vary wildly based on the model you use, which makes sense when you think about it but is not obvious when you’re first starting off. From creating embeddings to generating images, Matteo has you covered.
When building a microservice-oriented architecture, it can be difficult to keep track of and create a coherent feeling API. Depending on your implementation, you might have one API per service which can become impossible to keep track of for integrators. Consolidating your service APIs into a single gateway helps point consumers where they need to go simply and easily. Farzana Rahman shows us how to do that in her article on how to use a single base url for multiple services with API Gateway.
Matt Martz published an update to his CI/CD slackbot. This new change adds Amazon Bedrock into the workflow to give him summaries of what changed and identify any potential issues of the change going into a test environment. This is a really cool integration because not only does it analyze the code changes, but it also pushes the message into Slack so he can approve or deny incoming diffs with minimal interruption. I love seeing use cases for GenAI like this that extend beyond a simple chat bot.
Yan Cui has been building out his empire the past few weeks and creating really cool systems for pennies. Last week he published an article describing how he built web analytics reporting with Amazon Timestream. He goes into the reasons why he chose Timestream over DynamoDB and gives a high level overview of how it all works.
If you’ve ever built an API Gateway Lambda authorizer, you know how difficult they can be to debug. Luciano Mammino published an article last week about how he debugs Lambda authorizer issues. He walks through the steps necessary to get the logging you need and gives you IaC so you can do it yourself. I’m super happy I read this because it taught me about the
API Gateway Account resource that I didn’t know existed.
When AJ Stuyvenberg posts content on a serverless topic, it’s always well-thought-out and backed by tireless research. He recently turned his sights onto Lambda layers, publishing his findings in an article called “You Shouldn’t Use Lambda Layers.” He talks about the difficulties you encounter when using them extensively but also gives some use cases where they are a good thing. Fantastic read!
I’ve been on a bit of a real-time notification kick lately. Mainly I’ve been focused around how difficult it is to manage WebSockets. But Michael Walmsley might have changed my mind with his article last week. In his post about using AWS WebSockets without DynamoDB, he shows how you can start long-running tasks and report back to end users when it’s done simply by passing the connection id around in events. Not only does he show you how to do it, but he’s also made a CDK construct that does it all for you!
With re:Invent only 2 weeks away, we need to start preparing for all the contacts we’re inevitably going to be making. I really like Ben Pyle’s idea where he created a digital contact card. This could be super helpful to share info during a hectic week!
Getting ready for re:Invent I wanted to try something different this year to connect with everyone I meet. I'm going all in on a digital contact card. Trying it out so feel free to connect with me on any of these mediums. https://t.co/q2tbvRSc8B #HiHelloCard— Benjamen Pyle (@benjamenpyle) November 10, 2023
I don’t normally share updates to the AWS console because I don’t condone “ClickOps”, but I do like how the SQS and EventBridge Pipe integration makes it easy to tie AWS services together. I sometimes do things like this to figure out how to configure my IaC.
Your SQS to Lambda integrations just got faster. Polling from Lambda to SQS moved from 60 to 300 concurrent executions per minute, enabling faster processing of spiky workloads.
Lambda now reports timeouts and errors during the Init and Restore phases. Before you’d only get logs during the Invoke phase (when your code runs) but now you get pretty much the full lifecycle. Great update!
SQS now supports JSON protocol! Wow, it was a really big week for SQS, great job team!
I’m excited to meet as many of you as I can at re:Invent! If you want to meet up, please send me a message and we’ll get together!