The Distributed Transaction Problem

In a traditional monolithic application, maintaining data consistency is relatively straightforward when all related data lives in the same relational database. Suppose an e-commerce application needs to create an order and reduce the available inventory. Both operations can execute inside a single database transaction. If everything succeeds, the transaction commits. If something fails, the database rolls back all changes.

But what happens when we split the application into microservices? Now the Order Service, Inventory Service, and Payment Service each own separate databases. Creating an order requires coordinating changes across multiple independent systems. And unlike a single database transaction, there is no automatic mechanism that guarantees every service will succeed or fail together.

This is known as the Distributed Transaction Problem. It is one of the most important challenges in microservices architecture because it affects data consistency, reliability, and how business workflows recover from failures. Let's understand the problem, why traditional transactions become difficult, and how distributed systems handle it.