The article says <1sec for approximately 1B rows. what is the average data size used?

irrc each row is 5-10 string/int fields. With aggregate pushdown, there’s little overhead between Pinot and Presto.

Thanks. so if I am reading that article correctly which says 1B rows/sec for aggregation..so 1B * size of int = 4GB?

yeah that's the total size for each int column. Aggregation on 1b rows in 1 sec is achievable by Pinot. You can also refer to this other Pinot benchmark: https://www.slideshare.net/XIANGFU3/pinot-near-realtime-analytics-uber/18