We get sub-ms response time on a table with over 250 million rows. We have a covering index for all high-volume queries, running on rds (aurora) with 16gb ram.
Do you need to do aggregations/scans on the table as well or is it mostly direct row accesses?
Yeah, we do a lot of those and they do take a lot longer, and with much wider variance. On the order of 500ms-5s. These are all done async so it’s not a problem but we have begun dual writing to clickhouse to get us out of the aggregations business.