Agree. 20 million rows is nothing for even the most basic Postgres setup. However, at some point ClickHouse (or perhaps more dedicated ts databases) starts to make sense as the 23 byte per row overhead in Postgres weighs. Usually covering indexes are needed as well so it eventually becomes a little too much.
We were doing ok with about ~10B rows in Postgres before deciding to switch however. Even that might be fine for some workloads but not ours.
We were doing ok with about ~10B rows in Postgres before deciding to switch however. Even that might be fine for some workloads but not ours.