Why it matters

Hudi is streaming-first alternative. Understanding shapes streaming lakehouse.

Advertisement

The architecture

CoW: rewrite files on updates. Faster reads.

MoR: log deltas + async merge. Faster writes.

Hudi table typesCopy-on-writefaster readsMerge-on-readfaster writesIncremental querieschanges since XChoose CoW for read-heavy, MoR for write-heavy
Hudi trade-offs.
Advertisement

How it works end to end

Upserts: primary key-based. Handles CDC naturally.

Incremental queries: 'give me changes since commit X'.

Indexing: bloom index, HBase index, Bucket index.

Timeline: commit history.