Postgres as a Durable Workflow Engine: Benchmarking native statefulness vs separate orchestrators

Durable workflows can now run inside the DB. Postgres-backed statefulness is no gimmick—the game changer is DBOS, a Go library that automatically checkpoints a program’s state to Postgres. If the app crashes, workflows resume from the last completed step on restart ^[1].

Durable workflows without a separate orchestrator

With DBOS, there’s no separate service to run. It’s a pure Go package that auto-checkpoints into Postgres, so you can add durable workflows to an existing app without rearchitecting everything ^[1]. The result: reliability that kicks in automatically, even when failures occur mid-stream ^[1].

Why PostgreSQL goals matter

PostgreSQL's original project goals emphasized reliability and extensibility, and a Crunchy Data piece argues the creators nailed those aims ^[2]. That inheritance—trustworthy data, resilient design—shapes decisions about where and how durability lives: inside the database, not in a separate service.

Tuning notes for PostgreSQL 18

Tuning AIO in PostgreSQL 18 is relevant for in-DB workflows, with guidance that could influence performance as you shift more orchestration into the database layer ^[3].

Real-world use from the Go library

• AI agents handling unreliable APIs ^[1]
• Observable, fault-tolerant data pipelines ^[1]
• Orchestrating business processes with seamless recovery ^[1]
• Reliable background jobs with no timeouts ^[1]
• Processing incoming events (e.g., Kafka) exactly once ^[1]

Closing thought: this approach shows how the database itself can be the durable engine, but keep an eye on PostgreSQL 18 AIO tuning to squeeze more performance for in-DB workflows ^[3].

References