Nobody who got close to trying PgCache this year held off over speed. Mostly it was stale data, which is interesting, since lots of teams already read from a replica that lags the primary.
If that's you, here's a first step: test PgCache on the reads you send to a replica.

ALT A psql terminal window. The query SELECT replay_lag FROM pg_stat_replication returns one row: 00:00:00.284153. Below it, SQL comments read: "a replica 0.3 s behind: totally normal." "a cache fed by replication: wait, what about stale data?" "fair question. more in the post." Footer: PgCache, smart read replica for Postgres.
PgCache 0.8.0 is out!
@leonardobenedet, one of our design partners, benchmarked it the day after release. On a complex multi-tenant dashboard it served 24x the requests per second of Postgres alone at 100 tenants, and was still 1.5x ahead at 1,000 馃槑
We built PgCache but we've never met (yet!)
That'll change when we head to NYC in a couple weeks for Postgres Summit US/ @pgconfnyc
And, if you're in SoHo 9/30 to 10/2, and want to grab a coffee, send us DM --we'd love to connect!








