DynamoDB

Amazon DynamoDB provides serverless scaling for event sourcing on AWS with pay-per-request pricing.

---

Why DynamoDB

Strength	Benefit
Serverless	No capacity planning
Pay-per-use	Cost scales with traffic
Single-digit ms	Consistent low latency
AWS native	Integrates with Lambda, Kinesis
Global tables	Multi-region replication

---

Trade-offs

Concern	Consideration
Cost at scale	Can exceed provisioned alternatives
AWS lock-in	Not portable to other clouds
25 KB item limit	Large events need chunking

---

Configuration

[storage]
backend = "dynamo"

[storage.dynamo]
region = "us-east-1"
table_prefix = "angzarr"
# Optional: endpoint for local development
# endpoint = "http://localhost:8000"

Environment Variables

export AWS_REGION="us-east-1"
export DYNAMO_TABLE_PREFIX="angzarr"
export STORAGE_BACKEND="dynamo"
# AWS credentials via standard mechanisms

---

Table Schema

Events Table

Table: angzarr-events
Partition Key: pk (String) = "{domain}#{edition}#{root}"
Sort Key: sk (Number) = sequence

Attributes:
  - event_type: String
  - payload: Binary
  - correlation_id: String (optional)
  - created_at: String (ISO 8601)

Positions Table

Table: angzarr-positions
Partition Key: pk (String) = "{handler}#{domain}#{edition}#{root}"

Attributes:
  - sequence: Number
  - updated_at: String

Snapshots Table

Table: angzarr-snapshots
Partition Key: pk (String) = "{domain}#{edition}#{root}"

Attributes:
  - sequence: Number
  - state: Binary
  - created_at: String

---

Optimistic Concurrency

DynamoDB condition expressions enforce sequence ordering:

# Conditional put - fails if sequence exists
dynamodb.put_item(
    TableName="angzarr-events",
    Item={"pk": pk, "sk": sequence, ...},
    ConditionExpression="attribute_not_exists(sk)"
)

---

Local Development

Use DynamoDB Local for development:

# Start DynamoDB Local
docker run -p 8000:8000 amazon/dynamodb-local

# Configure endpoint
export DYNAMO_ENDPOINT="http://localhost:8000"

---

Table Creation

# Events table
aws dynamodb create-table \
  --table-name angzarr-events \
  --attribute-definitions \
    AttributeName=pk,AttributeType=S \
    AttributeName=sk,AttributeType=N \
  --key-schema \
    AttributeName=pk,KeyType=HASH \
    AttributeName=sk,KeyType=RANGE \
  --billing-mode PAY_PER_REQUEST

# Positions table
aws dynamodb create-table \
  --table-name angzarr-positions \
  --attribute-definitions \
    AttributeName=pk,AttributeType=S \
  --key-schema \
    AttributeName=pk,KeyType=HASH \
  --billing-mode PAY_PER_REQUEST

# Snapshots table
aws dynamodb create-table \
  --table-name angzarr-snapshots \
  --attribute-definitions \
    AttributeName=pk,AttributeType=S \
  --key-schema \
    AttributeName=pk,KeyType=HASH \
  --billing-mode PAY_PER_REQUEST

---

When to Use DynamoDB

• AWS native — Already on AWS
• Variable load — Traffic spikes/lulls
• Serverless — Lambda-based architecture
• Global — Multi-region requirements

---

Next Steps

• Bigtable — GCP equivalent
• PostgreSQL — Self-managed alternative