5 ms·
You can probably pre-compute part or all of those aggregates using triggers, and then run those queries in ~200ms or less with a plain B-tree.
by ddebernardy 10y ago
You can probably pre-compute part or all of those aggregates using triggers, and then run those queries in ~200ms or less with a plain B-tree.
- cldellow 10y agoYeah, that's what we did in the part we said "Pretend this is where we stopped" :) We pre-computed a few major dimensions and choose which table to query based on which columns are being requested. BRIN indexes were still a win vs B-tree, though. We keep being impressed by Postgres--we keep being able to flog it harder and harder instead of moving to something heavier duty.