View a markdown version of this page

Architecture details - Guidance for an Automotive Data Platform on AWS

Architecture details

DataZone V2 Domain

Domain: automotive-data-platform

See Platform foundation for the full deploy topology and stack naming conventions.

Configuration:

  • SSO: AWS IAM Identity Center (IDC)

  • VPC: Dedicated VPC with private subnets

  • Encryption: KMS customer-managed keys

  • Networking: VPC endpoints for AWS services

Projects:

  • 9 producer projects — one per data product (vehicle_telemetry_aggregated, customer_360, service_records, and 6 others; see Data products)

  • 1 smoke-test consumer project for subscription validation

Environments per project:

  • Development: Sandbox for experimentation

  • Staging: Pre-production testing

  • Production: Live data products

Data Product Registration

The foundation deploy automatically registers all 9 governed data products in the DataZone V2 domain. See Data products for complete schemas, partition schemes, and subscription patterns for each product.

Example: customer_360 Data Product

Registration (auto-provisioned by datazone-projects stack):

  • Technical name: customer_360

  • Owner: Customer domain producer project

  • Description: Unified customer profiles with snapshot-date partitioning

  • SLA: 99.9% availability, data freshness < 24 hours

  • Format: Apache Iceberg on S3

  • Access: Subscription request via DataZone V2 portal

Metadata:

  • Business glossary: Customer, Health Score, NPS, Churn Risk

  • Technical documentation: Schema definitions, query examples

  • Quality metrics: Completeness, accuracy, timeliness

  • Usage examples: Sample queries, downstream notebook templates

Example: vehicle_telemetry_aggregated Data Product

Registration (auto-provisioned by datazone-projects stack):

  • Technical name: vehicle_telemetry_aggregated

  • Owner: Automotive domain producer project

  • Description: Hourly-aggregated vehicle telemetry, partitioned by event_date with 16-bucket VIN bucketing

  • Format: Apache Iceberg on S3

  • Access: Subscription request via DataZone V2 portal

Metadata:

  • Business glossary: VIN, Sensor Reading, Diagnostic Code, Vehicle Health

  • Quality metrics: Completeness, accuracy, timeliness

Cross-Domain Data Sharing

Scenario: A consumer subscribes to both the customer_360 and service_records data products to enrich customer profiles with service history

Process:

  1. Consumer discovers the customer_360 and service_records data products in the DataZone V2 portal

  2. Consumer submits subscription requests via the DataZone V2 portal

  3. Producer domain owners approve requests (governance stack provisions Lake Formation grants automatically)

  4. Lake Formation grants column-level read permissions per the data product’s access policy

  5. Consumer queries both data products via Athena using the Glue Data Catalog

  6. DataZone V2 tracks lineage showing data flow between producer and consumer projects

Query Example:

SELECT c.customer_id, c.health_score, c.nps, s.service_month, s.repair_category FROM customer_360.customer_360 c JOIN service_records.service_records s ON c.customer_id = s.customer_id WHERE c.health_score < 50

Governance Policies

Data Classification:

  • Public: Aggregated metrics, no customer identifiers

  • Internal: Customer profiles with hashed IDs

  • Confidential: PII data, financial information

  • Restricted: Biometric data, health information

Access Policies:

  • Public: All authenticated users

  • Internal: Domain team members

  • Confidential: Authorized personnel with business justification

  • Restricted: Compliance approval required

Data Retention:

  • Transactional data: 7 years

  • Analytical data: 3 years

  • Logs and audit trails: 10 years

  • Temporary data: 90 days

Deployment Architecture

Infrastructure (see Platform foundation for complete topology):

  • Five per-stage CDK stacks: network, lake, datazone, datazone-projects, governance

  • One account-singular bootstrap stack: adp-shared-bootstrap

  • Resources: DataZone V2 domain, S3 Iceberg lake, Lake Formation tags, Macie session, CloudTrail trail, IAM Identity Center groups

  • Deployment time: 45–90 minutes via make deploy STAGE=<stage>

Integration:

  • All 9 data products registered automatically by the datazone-projects stack