Building a multi-tenant CRM with Amazon Aurora DSQL

This article demonstrates how to build a multi-tenant CRM system on Amazon Aurora DSQL using a pooled tenancy model with centralized tenant isolation and IAM authentication.

• Implement pooled multi-tenancy with tenant_id on every row as the isolation discriminator
• Use verified token claims for tenant resolution, never client-supplied headers
• Leverage IAM-based authentication with short-lived tokens instead of database passwords
• Apply Optimistic Concurrency Control (OCC) for concurrent writes without locks
• Batch large operations to respect transaction limits (3,000 rows, 10 MiB, 5 minutes)
• Use asynchronous index creation with CREATE INDEX ASYNC for non-blocking background builds
• Combine foreign keys with application-level tenant ownership checks for referential integrity
• Deploy on Amazon ECS Express Mode with automatic scaling paired to Aurora DSQL serverless architecture

The design patterns—verified tenant context, tenant-scoped repositories, and OCC retries—apply to any transactional multi-tenant workload on distributed SQL databases.