I like to start this newsletter with some stats. This blog post that summarize a paper published by Microsoft, in which MS provides statistics about Azure Functions usage. An interesting number is the average execution time. Around 20% of the functions run in less than 100 ms and about 50% run in less than 1sec. If you remember the report from Datadog about Lambda, it was about the same.

Let's continue with Azure... The article goes throughout the creation and use of custom handlers in Azure Functions. Custom handlers allow to run your own command, script or executable, even in a language not supported by Azure Functions.

A really complete tutorial by Sean O'Toole from Capital One where he gives 11 best practices to create lambda functions with Java. As a former java developer, I'm happy to see this kind of article, even if I personally moved to python when I code lambda functions. Among other recommendations: reduce your dependencies, avoid big dependency injection frameworks (Spring), avoid reflection, …

2-parts tutorial on how to build a pretty standard serverless architecture (API GW, Lambda, Dynamo, Cognito), using Java (yes, again!). The interesting part is that the author, also give feedbacks on his experience. Good articles for people who want to start.

Really complete article that details every part of a typical serverless architecture. Don't be scared by the big diagram, serverless is there to make things easier, and Xavier Lefèvre explains every part of it. A must read!

Good blog post from James Beswick, as usual, in which he talks about how to organize code in your repo, how to reduce the amount of code by leveraging the appropriate serverless service and how to deploy it using CI/CD and canary deployments.

In this article, Richard Fan details how to leverage API Gateway to avoid using Lambda functions in some case. Definitely, it seems a hot topic and definitely something you should evaluate in some cases. Actually, if you missed my talk on Tuesday, the topic is the same with a slightly different approach.

I really love CDK, the AWS Cloud Development Kit, and building infrastructure as real code (python, js, java…). In this AWS blog post, the author demonstrate a simple architecture (API Gateway, Lambda, Aurora Serverless) and how to build it with CDK and typescript.
If this topic interests you, you should visit the CDK pattern website and have a look at a recent presentation of CDK by Darko Mesaros.

In this article, Vittorio Nardone describes how to test lambda functions locally with Docker, leveraging the lambci images. I would rather recommend using sam (sam local invoke) which also use the lambci images but it gives an idea of what happens under the hood.

In this article, Arnaud Charlier talks about the dark side of serverless. As he says, "serverless = limitless" (or quite). So if you do things wrong (in his example, an infinite loop with S3 events and Lambda), it can be very costly (especially Cloudwatch logs). A good practice (among others) is to set a budget and alarms, so you know when something goes wrong.

In this article, Rebeca Chinicz describes step-by-step how to build a small e-commerce application. Everything is here, the website in S3, cached by Cloudfront, the API exposed with API Gateway, the authentication through Cognito, the backend with Lambda and DynamoDB as the data store. A pretty standard serverless architecture and another good tutorial.

Let's conclude this newsletter with a bit of fun. As a former MTG player, I really like this one! How they built an Online Card Game based on AWS Lambda and Google Firestore in one week. Yes you read correctly, a mix between AWS and Google.