How one can disable cache or preaggregation for large datasets?

**Problem**

Sometimes users can run smth like 'Select many dimensions and metrics for long period ".
When it happens, huge portion of data is added to memory cache (we are still using redis, but I think problem will remain after adding cube store).

How can we disable adding result of query to cache by its size (e.g. limit 30 mb)? For us it's cheaper to make roundrip to db)