AWS SAM

Recently I used AWS SAM on a job and in my pet project. I haven’t worked on the backend side of things for a while and it was a nice way to create infrastructure compared to things I was dealing with before. In this blurb I’m going to describe my early experience with creating infrastructure and writing some business logic.

What is it?

AWS SAM lets you declaratively describe infrastructure.

Resource types

SAM lets you create a bunch of resources that will cover needs most of the web apps:

Lambda for running code
Databases
File storage
Queue
CDN

1Resources:
2  <ArbitraryId>:
3    Type: AWS::<Known Resource Type>
4    Properties:
5      DifrentResources: “Have different properties”

Serverless functions

Lambda is the most important building brick that runs the business logic and becomes a glue between other resources. For me, the definition of a lambda usually looks like this:

1Type: AWS::Serverless::Function
2Properties:
3  Handler: src/my-fun.handler
4  Runtime: nodejs14.x

And the src/my-fun.js would look something like this

1exports.handler = async (event, context) => { }

Triggers

Lambda can be used in various ways:

Creating Rest API
Handling Queue messages
Handling CDN redirects
Reacting to file uploads
Reacting to table row changes

Rest API

1Events:
2  <ArbitraryId>:
3    Type: Api
4    Properties:
5      Path: /<path>
6      Method: post

Few lines of config lets you add a REST API method. Behind the scenes it creates API Gateway that configured to call your Lambda function. Now your web client or other services can invoke the function using HTTP and get a reply.

Queue

1Events:
2  <ArbitraryId>:
3    Type: SQS
4    Properties:
5      Queue: !Ref MyQueue

This means that your function will be invoked with a message that came to MyQueue queue. It’s helpful for creating loosely coupled systems and for handling jobs that take too much time to be handled during a REST request.

CDN

1<ArbitraryId>:
2  Type: AWS::CloudFront::Distribution
3  Properties:
4    DistributionConfig:
5      DefaultCacheBehavior:
6        LambdaFunctionAssociations:
7          - EventType: viewer-request
8            LambdaFunctionARN: !Ref RedirectLambdaFunction.Version

You can configure CloudFront to call your lambda in case of a file missing in S3 bucket. This is useful for creating files lazily. Note, that this is not Event property but rather configured on the CloudFront config.

S3

1Events:
2  <ArbitraryId>:
3    Type: S3
4    Properties:
5      Bucket: !Ref <BucketId>
6      Events: s3:ObjectCreated:*

You should rather rely on S3 triggering your function upon upload completion than your web client. This is useful when you need to create a database entry for the uploaded file or you need to process the file.

DynamoDB

1Events:
2  <ArbitraryId>:
3    Type: DynamoDB
4    Properties:
5      Stream:
6        "Fn::GetAtt":
7          - <TableId>
8          - StreamArn
9      StartingPosition: TRIM_HORIZON
10      BatchSize: 100

Triggering a Lambda function on changing DynamoDB table might be useful for creating loosely coupled system but also I saw a case when it was used in conjunction with TTL in order to calculate a new state of an entity.

Things I missed

It’s easy to start using Lambda and quickly create business value. Having said that I would love to see some improvements that will help developers create modern scalable solutions from the get go.

Native Typescript

I have used javascript for my pet project but for the system in anger I used Typescript. Typescript lets you write scalable code that can be supported by future teams. At the time of implementing my projects, there wasn’t an official approach to this. Recently, AWS published a blog post that describes the new built-in way of doing it. If I were you, I'd start there.

But if you want to follow your own way, you can transpile Typescript using bundlers like Webpack. In particular, I was using this tutorial. I like that this approach only uses public SAM API and doesn’t lock all your Lambdas to Typescript.

Also, there are projects, rather opinionated ones, that let you use one command to build everything like this one.

Not the latest Node.js version

One of the surprising things for me was the fact that AWS Lambda doesn’t support the latest version of NodeJS but only 14th and 10th versions. This is not a big deal but it caught me off guard when my code failed because of usage of newer ECMA syntax.

Database

Every app needs to have a way to store data permanently. Luckily AWS SAM lets you describe infra resources to store data in relational database, as well as NoSQL solution.

NoSQL – DynamoDB

1<ArbitraryId>:
2  Type: AWS::Serverless::SimpleTable
3  Properties:
4    PrimaryKey:
5      Name: <ArbitraryId>
6      Type: String

Used it a couple of times. Cheap & fast.

File Storage

1<Name>:
2  Type: AWS::S3::Bucket
3  Properties:
4    BucketName: <ArbitraryIdIfNeeded>
5    CorsConfiguration:
6      CorsRules:
7        - AllowedHeaders:
8            - "*"
9          AllowedMethods:
10            - GET
11            - PUT
12            - POST
13          AllowedOrigins:
14            - "*"
15          MaxAge: "3600"

Simple Storage Service is a service that I used on every project I that was powered by AWS. Not surprisingly, in both projects users were uploading media files and a lambda was processing them.

CDN

1<ArbitraryId>:
2  Type: AWS::CloudFront::Distribution
3  Properties:

Then, you can create a subdomain for the bucket that is human-readable (e.g. files.marsel.name) and loosely coupled (e.g. you can point it to Azure Blob Storage in the future).

RDS

Even though I like SQL for the ability to quickly run queries for data analysis, I haven’t used any of the SQL databases. While researching for this article, I found out that there is not a simple way to create an RDS instance using AWS SAM. What a surprise! Check these out first if you’re interested in creating Aurora instance – 1, 2.

Things I haven’t used

There are many capabilities that I haven't used. In particular SQS, SNS, GraphQL, Step Functions.

Examples

If you ever get stuck, google it or search your case in the following resources:

Alternatives

Terraform

It supports several cloud providers but a bit wordy and complex for me.

Architect

Haven’t used it but heard good things.

Conclusion

I used it briefly to construct some of the parts of larger systems but I think AWS SAM is a great tool to run the whole project’s infrastructure.