April 30, 2026
Hot WAL summer
Does Postgres Scale?
Postgres flexes huge numbers, but the comments instantly turned into a database cage match
TLDR: The article says one powerful Postgres server can handle surprisingly huge app workloads, enough for most real-world companies. In the comments, fans called that more than sufficient, while skeptics argued Postgres only shines if you know a bag of weird tuning tricks.
A benchmark claiming Postgres can absolutely handle the heat sent the community straight into debate mode. The big brag: one beefy cloud server reportedly pushed 144,000 writes per second and handled 43,000 workflows per second—basically saying this old favorite can power a mountain of app activity without breaking a sweat. The article’s core message was simple: for most companies, a single Postgres setup is probably way more than enough.
But the real entertainment was in the replies, where the vibe split into "calm down, this is plenty" versus "sure, if you know the secret spells". One commenter basically said most teams panic about giant scale way too early, acting like they’ll become the next mega-platform any minute now. Another came in with the grumpy veteran energy: yes, Postgres scales, but only after you learn a pile of weird tricks and undocumented magic settings. That comment hit a nerve, because it turned the story from a victory lap into a familiar tech soap opera about tools that are "easy" until they aren’t.
Then came the product fandom. One fan hyped DBOS as the simpler rival to Temporal, dropping the killer analogy that Temporal is like Kubernetes while DBOS is like docker compose—which is exactly the kind of nerdy shade the internet lives for. And for extra spice, someone tossed in a link like a mic drop, reminding everyone that the "can Postgres scale?" discourse is now basically its own long-running franchise.
Key Points
- •The article benchmarks a single PostgreSQL server to measure write scalability for durable workflow workloads.
- •In the point-write test, the server sustained up to 144,000 inserts per second, which the article equates to about 12 billion writes per day.
- •The benchmark environment was an AWS RDS db.m7i.24xlarge instance with 96 vCPUs, 384 GB RAM, and 120K provisioned IOPS on io2 storage.
- •The reported bottleneck was WAL flushing to disk, with one process performing group commit flushes while many others waited on the WAL lock.
- •For simple no-op durable workflows requiring two database writes, the article reports throughput of up to 43,000 workflows per second, or about 4 billion per day.