How do I process billions of records in ADLS data using spark less time

Question

I have data that has avg of 1-3billion rows per month. I am doing simple sql query to filter out data basis date(ie part of partition) and few more conditions. But this takes 6-