• Inheritance: In the OO world a User object might serve as a base class for users of type teacher or student. How do you map this relationship in the relational database?
• Granularity: This happens if the classes do not match up to the tables in the database. You might for example have a situation where a User is associated to a country. In your database you might represent this as two tables, but in your class you might represent the country as a property or field of the User class. So there exists a mismatch. How do you map this situation?
• Identity and Equality: Lets say you query a database for a User with ID 100. The result you get back will always point to user with ID 100. If you look at it from an OO perspective, will a User with ID point to the same object, or multiple instances of it?
• Association (directionality): In an OO model you associate classes with each other by placing references in each class of the other. In a relational model you create the associations by placing a foreign key in a table.
• Type systems: The types in a database are quite different from those in an OO environment. For instance a VARCHAR with a size of 100 is not the same as a string, because a string is dependant on the amount of memory available.

I then read two StackOverflow topics, one here and one here. From what I could gather ORM’s are not great at performing bulk operations, but what they lack in that area they make up for in developer productivity.

I remember coming across an article that compared the performance between various ORM mappers on this website.