Firebase is great for building real-time apps, and it uses a tree topology to model data. Fundamentally, we are using a graph topology to model data which has some interesting upsides, for instance, data denormalization is not needed.

I read that the reasons they choose a tree structure is because of the speed. A tree structure with no cycles make sense to have a better speed. How do you compare this issues in firebase and appbase?

Imo, latency difference shouldn't be huge between a tree/graph data topology. Graph traversals for our read requests takes less than 10ms.