Episode 121: OR Mappers with Michael Plöd

Performance

The performance of Object-Relational Mappers (OR Mappers) is often debated, but Michael Plöd argues that they generally perform well compared to hand-coded SQL, especially when used by average developers. He notes that while database experts can outperform OR Mappers, these tools offer abstraction and portability that hand-coded SQL cannot 1. Plöd compares this to the performance impact of virtual machines versus assembly language, emphasizing that OR Mappers do not inherently degrade performance when used correctly 2.

You will always find people that say OR Mappers are so slow. But I think the reality is that not every developer can be a database expert in terms of writing the top performing SQL for database x.

--- Michael Plöd

He also highlights that a good OR Mapper sends efficient SQL to the database, reducing the need for developers to manage low-level performance details 2.

   

Caching

Caching in OR Mappers is a nuanced topic, with Michael Plöd advising a conservative approach. He suggests tuning data access before implementing caching, as not all data benefits equally from it 3. Plöd warns against caching critical data, like financial information, due to the risks of outdated caches, while recommending caching for static data like country lists 3.

Before I start introducing a second level caching mechanism to an OR Mapper, my recommendation is to first tune the plain data access.

--- Michael Plöd

He also stresses that caching should only be added when it provides business value, and not just because it seems like a trendy solution 4.