Start with the access pattern
Row and column storage are different trade-offs, not a simple ranking. Row layout fits processing a complete record together, while column scans can read only the fields a query needs and compress similar values effectively.
Once this is clear, file formats and scan operators in ClickHouse, Doris, and Spark become easier to reason about.