While large monoliths have their transaction boundaries well-defined, and because we often use techniques such as two-phase commits when we are writing microservices, we have to approach these requirements in another way.

We should remember that each microservice is the owner of its own data storage, and we should access their data using only their APIs. It is important to keep your data up to date and use compensating transactions when an operation doesn't work as expected. When we write monolithic applications, many operations are executed as a single transaction. For microservices, we need to rethink the operations and transactions to make them fit within each microservice boundary.