Oracle has a built-in read cache .. why doesn't PostgreSQL? Teams migrating from Oracle end up rebuilding it in the app. At PgCache we think there's a better way to serve reads for Postgres: a drop-in proxy that automatically caches hot reads. pgcache.com/try

Reliability is indeed one of the big reasons that James is building @PgCache on PostgreSQL!

@PgCache won't make it to Valencia - but we'll see you next year!

transactions now supported 👇

Here's the free WASM tool we built to speed up some of the analysis for you: pgcache.com/fit

Which Postgres queries are worth caching? We shared a free online Postgres tool on Reddit yesterday. 26k views and 180 shares so far! The idea behind it is pretty simple: rank your queries by total time vs what "feels slow."

If you're mid-decision on a cache right now, I'd love to hear what pushed you there.

Want to know if your workload fits before you build anything? Paste your pg_stat_statements into the Fit Analyzer. It shows which queries PgCache would cache, which pass through, and why. Runs in your browser: no signup, nothing uploaded. pgcache.com/fit

Under the hood it's a Smart Read Replica: it holds only the data your traffic touches and keeps it current from the logical replication stream. Writes and anything it can't cache go straight to Postgres. (Not a fit yet if your reads rely on RLS.)

Several teams have come to us lately at this exact point: cache decided, not built yet. What they wanted to skip: cache logic around every hot query, key deletes on every write path, TTLs to second-guess in production. That's what we built PgCache for.

Already decided Postgres needs a cache? You don't have to build it into your app. PgCache sits between your app and Postgres and caches your hot reads. Swap the connection string. No cache code, no TTL to guess at.

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!

Don't wonder if caching might help. Find out. Takes < 1 min (locally in browser via WebAssembly). Check it out at pgcache[.]com/fit

You probably don't need another read replica. What you probably need is the answer to this question: "how much of our read traffic can be cached?" Check yours for free in 2 minutes, locally in your browser, at pgcache [dot] com/fit.

We see a third path for scaling reads on PostgreSQL Path 1 - Read replicas & server upgrades. Path 2 - External caches. Path 3 - CDC-driven smart read replicas. We saw what's possible with PostgreSQL Logical Replication, and thought "it doesn't have to be this complicated."

@HuluWZ built an independent PgCache benchmark with a custom ecomm workload and some slick design choices. No diff on point lookup speed (not our focus)... But he saw 2.9 seconds -> 0.6 ms on a 10M-row join 🔥 dev.to/huluwz/how-i-put-pgca…

Independent benchmark of PgCache. 16M rows, published in full. No difference for point lookups, because that's not what we're solving for (0.3ms → 0.3ms). But the 10M-row join? 2.9s → 0.6ms Unpaid, unaffiliated, MIT repo: dev.to/huluwz/how-i-put-pgca…

In the age of Logical Replication there's simply no reason a human or agent should have to manually configure invalidation, TTLs ... much less decide which queries get cached. PgCache does it all for you. Check out our benchmark harness. github.com/PgCache/pgcache-b…