MongoDB：比较两个大集的最快方法

I have a MongoDB collection with more than 20 millions documents (and growing fast). Some document have an 'user_id' (others, don't).I have a MongoDB collection with more than 20 m