Let's start this newsletter with a story: the story of Payam Moghaddam from Galvanize about their journey to Serverless. By moving to serverless, they've been able to focus more on their business. But serverless is not magic and it's actually a journey for most people. This article describes this journey, the frictions you may have (with code organization, dev|ops boundaries, testing, ...) and how to solve them.

Who said Java was not a good idea for Lambda functions? OK, you may have cold starts longer than for other languages, but if you read this article and this other one from "I love my local farmer", you will have many recommendations to reduce these cold starts. You don't code Lambda functions like you build microservices: forget about Spring Boot, take care of your dependencies and your final package size, initialize your code properly (outside the handler function), ...
To continue with My Local Farmer, 2 other articles have been posted recently about Database connection from your Java Lambda function: part 1 & part 2, using RDS Proxy.

Everyone wants to become the next Salesforce, Square or Shopify and to provide a great SaaS (Software as a Service). But to achieve this, you will need a robust multi-tenant architecture. In this article Sena Yakut explains the different approaches to do it with Cognito: 1 User Pool per tenant, using groups or custom attributes. This other article ("Serverless architecture for multi-tenant SaaS") touches upon another topic: costs, and how to keep costs low while maintaining strong data isolation between tenants.

This is a topic I really enjoy, and I'm happy to see an AWS blog post about it. In this article, Luca Mezzalira describes why and how to apply hexagonal architecture in a Lambda function. One of the main reasons is evolutivity, a nice way to counter "vendor lockin". The other reason is testability.
Testability, which is the first point of many others best practices listed in this other article: "AWS Lambda Functions Best Practices".

Have you ever seen a line of Cobol code? Me neither! At least until I read this article (2 parts) on the AWS blog, explaining how to actually deploy a Lambda function written in Cobol. And if you wanted to recycle some of your mainframe code, now you know how!

You feel this newsletter is for dummies? Then I have something for you: this video from Marc Brooker. You probably know you can deploy functions as container images since last december, and these images can be up to 10GB. But then, how to quickly start functions? Listen to Marc and you will know. Really, really, really deep dive!

Last month, AWS announced the AWS Step Functions Workflow Studio to ease creation of Step Functions state machines. But it remains complicated to test these state machines. In this article, Yan Cui provides his strategy and some insights to test workflows.

You probably know Gregor Hohpe for his articles on architect[s|ure] ("The Architect Elevator") or for one of his books ("Enterprise Architecture Patterns"), probably less for his development skills. In this (2-part) article, he tries (and succeeds) to develop a loan broker using Step Functions, Lambda and other serverless services. It's actually really well detailed for an enterprise strategist...

Let's continue with Step Functions. Great story and details on how IHS Markit (Financial company) was able to reconcile hundreds of reports using Step Functions. Definitely a service I love!

When talking about serverless, we often think "Lambda". But sometimes Lambda does not fit business requirements, especially when latency is an important criteria (cold starts can be problematic if the load on your function is not consistent). In this article, James Eastham explains how he migrated from Lambda to AWS App Runner, a recently announced serverless container service.

Last year, Julian Wood (AWS Serverless Developer Advocate) started a series of articles on well-architected serverless applications. One year later, the series continues with new articles on security and reliability:

• 5.1. Resource policies, controlling and blocking network access, and audit tools
• 5.2. Using temporary credentials, smaller single-purpose functions, and auditing permissions
• 6.1. Security awareness documentation, inspecting code dependencies, and validating inbound events
• 6.2. Securely storing, auditing, and rotating secrets used in application code
• 7.1. Throttling and performance testing
• 7.2. API quotas and mechanisms to protect non-scalable resources

So you've built an API on AWS and you would like to use it securely from onprem applications? This AWS blog post (one more!) goes through the details of the VPN, VPC endpoint and DNS configuration to achieve this. Not an easy task (at least for me), so definitely useful.

Lambda is quite easy to use... but this simplicity hides the complexity under the hood, the one you don't really have to care about. But it's really interesting to see how things are actually working: scaling, load balancing, asynchronous invocation, firecracker...

Let's finish this newsletter with a song from Jeremy Daly, author of off-by-non newsletter and serverless chats podcast. "The answer that's best, to clean up all that mess, the answer is goin' serverless 🎶 ". See you in September!

AWS updates

AWS SAM Pipelines. With this new SAM CLI feature, it will be easier to create CI/CD pipelines for your serverless applications. And the great news is that it also works with Github Actions, Gitlab CI/CD and Jenkins (in addition to CodePipeline). Cool stuff!