The bulk of the overhead will be based on data; having lots of resource types and relations should be fine, although you might want a more dynamic model if you have that many