Episode 6. RocksDB and LSM trees. Day 6 of your 12 day sprint.

Scenario: "When would you pick RocksDB over Postgres?" Answer with the write path, not the brand.

Say: "Writes land in a memtable backed by a WAL, freeze into sorted SSTs, and compact down through levels. Reads merge across levels."

Say the three amplifications: write amplification, read amplification, space amplification. Say: "Compaction trades write cost now for read cost later."

Say the strategy line: "Leveled compaction favors reads. Universal favors write bursts. There is no free setting."

Lab: read one RocksDB wiki page on compaction, pick two tuning knobs, and narrate each knob's trade-off aloud for sixty seconds.

Depth tag: Recognize. Say: "I read LSM structure and narrate the method. I have not implemented compaction code."

Drill: define all three amplifications aloud: write, read, space.
Say: memtable to SST to levels, and every level merge has a price.
Say the strategy line again: leveled for reads, universal for write bursts.
Done-check: narrate two tuning knobs and their trade-offs in sixty seconds each.
Honest close: method narration is the claim, implementation is not.

Teaser: tomorrow, Episode 7, consensus. Raft, quorum math, and why minorities must stop writing.