We have an opportunity to reduce our MongoDB memory usage. I discovered recently that our base Model class indexes the id field as a unique index. The id field is defaulted to a string representation of the _id field (which probably isn't terribly useful). Since _id must always be indexed in MongoDB, we essentially index that same information two different ways (which uses more RAM).

I am not sure how many models this affects, but I did find at least one: the CallResource Model. It doesn't override the id field, and it does not override the unique_indexes class attribute.