Once upon a time... Let's start with a story, the story of Babbel, the famous language learning platform. Babbel has been on AWS since 2007, and needed to update their stack, we can name it a legacy (EC2, Opswork and Chef). In this article, a senior manager from Babbel's infrastructure team explains why and how they migrated to serverless technologies, mainly ECS Fargate and Lambda, and what were the benefits.

In this article, Jason Lengstorf compares two approaches for front-end developers to deploy their application (actually not only the front end but also APIs and others). One is using Node, Express and deploying it on Kubernetes. The other is using functions as a service, here Netlify. And the winner is... guess?

Excellent tutorial from Nader Dabit on creating a full-stack serverless application, with react for the frontend and Appsync for the backend (GrahpQL). This article is autonomous but if interested there are 2 other parts before, explaining serverless and Amplify.

When using API Gateway, we often think about REST or HTTP APIs, but there are also Websockets. If you need to communicate with the client, in real-time, without implementing a plain old polling mechanism, you can have a look at this article that goes through the implementation of such an API.

Did you know there are actually servers behind "serverless" 😁 ? And if you want to know more about the underlying architecture of Lambda, this dive deep article will give you all the insights: workers, Firecracker, load balancing and autoscaling, async vs sync invocations... you'll know (almost) everything!

I've shared this kind of article already in the previous newsletter, but if it can save you hours of decision making, it is beneficial. Great feedback from Tobias Schmidt on many different topics: layers, EFS integration, Lambda@Edge, concurrency, security, infra as code....

Modernizing, again and again, to benefit from the new capabilities the cloud offers. This is the topic of this article from Sheen Brisals. He doesn't speak about modernizing, but refactoring, and why you should continuously refactor your application, especially in the serverless world where things evolve quickly.

CI/CD is a key piece in software development and no exception for serverless. And Github is one of the leaders for managing source code and automating build and deployment actions. Recently, AWS released the "setup-sam" Github action so that you can install SAM and then use it to build and deploy your serverless applications described with SAM templates. Mandatory if you are on Github!

One other pattern from Devops practices is the use of feature flags, so you can decouple business deployments from technical deployments, perform A/B testing and easily rollback without redeploying. This article describes how you can do that with Lambda, AppConfig and the AppConfig Extension for Lambda. Neat!
By the way, I take the opportunity to quickly share a lib I contribute to: FF4J (Feature Flipping for Java). I've been working on the DynamoDB and Parameter Store repositories. Have a look if you code in Java!

Idempotency is a key pattern when working with Lambda. As your function can be retried several times, you need to ensure it won't perform the same operation several times. This article gives a simple example with Step Functions.

Really great blog post series on the AWS blog, from Benjamin Smith. 5 parts (until now) to learn how to build serverless applications:

• Part 1: Introduction
• Part 2: Business Logic, using AWS Lambda
• Part 3: The front door with API Gateway
• Part 4: Local development workflow
• Part 5: Sandbox developer account

When using serverless technologies, and Lambda obviously, you can quickly end up with a hundred functions (if not more). Observability is key to understanding what happens and to troubleshoot. This article gives interesting recommendations about logging: correlate your logs (using correlation ids), write logs as JSON to get more power when analysing them and be mindful of excessive logging as Cloudwatch can quickly become the main source of expenses.

I really enjoy this kind of story where you need to process hundreds of thousands items and you actually do it with Lambda. Great feedback from Mohammed Lutfalla on the challenges he faced and how he resolved them.

A bit of fun to close this newsletter. Could your wife (or any non-IT person) code a Lambda function in 7 days? This guy did the experiment and his wife was able to do it. Now ask the same with a Kubernetes cluster...

AWS updates

One update actually made my month: AWS Step Functions Workflow Studio. With this new feature you can now visually create your state machine. This is not the end of ASL (Amazon State Language), but it will certainly speed up the creation of workflows. You can also have a look at this blog post to have a concrete example. And I beat my own record on LinkedIn with this news (more than 15000 views, almost 300 likes), crazy!