We have a large MongoDB collection, 6TB and growing a lot. The collection is used for user and automated feedback, and as such will be used for all sorts of analytics. One document has about 20 fields. Since collection is used for analytics, filtered fields can be of any combination and as such, creating indexes for every combination would be too much.

Is there a good way to index collection for such (random-like) queries? If not, is there a way to optimize queries themselves?

If this can help with optimizing, queries are paged.

Example query:

db.usage.find({"appVersion": "4.0.0.0", "expression": "abcd"})

where appVersion is not indexed.