The principle of Peristsence Ignorance (or PI) is that our applications and Domain should be unaware of how, or even if, their Aggregates are being persisted.
By adhering to PI we ensure that we achieve a clear decoupling between our Domain and the mechanism we are using to store and retrieve our data - be this a database, a REST layer, web services in an SOA, an in memory database, or anything else.
To achieve this in DDD we use the Repository pattern.
Blog post: Blogged There Is No Database