[jpgvm]

If you are doing this sort of work I highly recommend the Datasketches library. It's used in Druid, which is a database specifically designed for these sorts of aggregations over ridiculously large datasets.