Distributed Data Management

• Each microservice should have its own data persistence layer
• distributed microservice architectures trade consistency for performance
• need to embrace eventual consistency
• you end up with some partial transactions, so we need some control logic to redo already processed transactions

Distributed Saga Pattern

• In the case of a failed transaction Saga orchestrates a series of compensating transactions that undo the changes that were made
• AWS Step Functions make it easy to implement a Saga execution coordinator