My current data model has a tree like structure, but I need to have primary keys on every tree level. So something like this:

+tenant
|
+--+tenant_region
 |
 +---+ client_group
 |
 +---+client

The docs recommend to use joining via edgeCollections in such a case. The driver api seem to forces named graphes as written in the docs.

How can I create and join(query) edgeCollections without creating a graph as recommended in the cookbook?

[EDIT]
The docs tell us that anonymous graphs are faster (I guess only at writing stuff to the db?). How much faster is faster? Maybe I worry too much about performance?