Serverless at scale on AWS: three Terraform + Lambda scenarios

AwsSamples is sample code for large-scale web sites on AWS, organised as three self-contained scenarios — each with its own Terraform state, deploy script and README:

• shopping/ — high-scale e-commerce: catalog, cart, checkout, fulfillment on CloudFront + WAF, S3, API Gateway with Cognito JWT auth, Lambda, DynamoDB, SQS + dead-letter queues, Step Functions and EventBridge.
• gaming/ — a realtime game backend: matchmaking, live events, leaderboards and telemetry over API Gateway WebSocket, Lambda, DynamoDB and Kinesis.
• chat/ — large-scale WebSocket chat with SQS fan-out and a browser client.

The mental model every scenario shares

• Stateless compute everywhere. No EC2, no containers — Lambda scales without an operator. Handlers are plain .NET 8.
• Managed data planes. DynamoDB with pay-per-request billing, TTLs and global secondary indexes instead of servers you patch.
• Async by default. SQS queues with DLQs buffer spikes; Step Functions choreograph multi-step flows; EventBridge fans events out.
• One deployable per scenario. publish.ps1 zips the Lambda handlers; terraform apply creates everything else.

Deploying one yourself

cd shopping
./lambdas/publish.ps1        # dotnet publish + zip each handler
cd terraform
terraform init
terraform plan -out plan.tfplan
terraform apply plan.tfplan

Prerequisites are Terraform 1.6+, the .NET 8 SDK and the AWS CLI. outputs.tf prints the endpoints when it finishes, and terraform destroy removes everything — which makes this a safe sandbox for learning.

Why single-region on purpose

The samples stay single-region so they stay readable; the README's hardening notes sketch the multi-region path (Route 53 failover, DynamoDB global tables, per-region state). Read the Terraform files top to bottom — they double as an architecture course, the same way the Azure Terraform article does for Microsoft's cloud.

Repository: github.com/bobhuang1/AwsSamples