DBOS proposes Postgres-backed durable workflow execution
Original: Building durable workflows on Postgres
Why This Matters
Simplifies distributed workflow architecture by eliminating external orchestrators
DBOS argues that Postgres can replace external orchestration systems like Temporal and Airflow for durable workflows. Their approach uses Postgres directly as coordinator, with application servers polling workflow tables and checkpointing progress to the database instead of using separate orchestrator servers.
DBOS proposes using Postgres directly for durable workflow execution instead of external orchestration systems like Temporal, Airflow, and AWS Step Functions. In their model, application servers communicate directly with Postgres rather than through a central orchestrator. Clients submit workflows by creating entries in a Postgres workflows table, while application servers poll the table to dequeue and execute workflows. Progress is checkpointed directly to Postgres, enabling recovery if servers crash. The company argues this eliminates the complexity of separate orchestrator servers, as application servers can coordinate through Postgres using mechanisms like locking clauses to ensure each workflow is processed by exactly one worker. DBOS offers open-source libraries for TypeScript, Python, Go, and Java, plus a cloud deployment option.