Besides modeling entities and their relationships, a key design choice in terms of consistency guarantees that the persistence layer (database) needs to give to the application. If a use case involves modifications of two or more entities, these guarantees from the storage system play a pivotal role in system architecture and SLAs. For example, consider the account transfer use case for a Banking application. After the transfer is done, the net debit/credit amounts should tally—no matter what happens in terms of infrastructure failure. Such logical units of work (debit from account x and credit to account y) are called transactions.
Let's look at some guarantees that databases (and storage systems in general) provide for transactions.
