Building a Lakehouse with DuckDB and Iceberg

Why open table formats changed the calculus

For years a lakehouse implied a cluster. Open table formats moved the transactional guarantees into the storage layer, which means a single well-provisioned node can now handle workloads that used to justify a distributed engine.

What the setup looks like

Land raw files in object storage, register them as tables, and let the query engine handle snapshot isolation. Compaction runs on a schedule rather than as an afterthought, and time travel becomes a debugging tool instead of a marketing bullet.

The result is a stack with fewer moving parts, a far smaller bill, and a mental model your whole team can hold in their head.