Columnstore indexes are best suited for which workload?

• A. Online transactional processing with frequent updates
• B. Small, random-access reads
• C. Analytics workloads with large scans
• D. Real-time row-level updates

Answer: (C)

Columnstore indexes organize data by column, not by row, which makes them extremely efficient for analytics that scan large portions of a table and perform aggregations across many rows. Reading only the necessary columns reduces I/O, and the data is highly compressed, so even more data can be processed with less disk and memory traffic. In addition, batch mode execution processes many rows at once in a vectorized way, and features like segment or partition elimination let the engine skip irrelevant data, speeding up long scans. This combination is why analytics workloads with large scans see the biggest benefits.

For contrast, online transactional processing workloads rely on fast, frequent updates and quick single-row lookups, which are better served by row-oriented storage. Small, random-access reads target specific rows where a columnstore’s layout would lead to unnecessary reads of entire columns. Real-time row-level updates similarly require quick, row-by-row access, which columnstore isn’t optimized for.