What is the role of vectorized execution in modern data engines?

Vectorized execution processes batches of data at once instead of row-by-row processing.

Vectorized engines process data in columnar batches using CPU-efficient operations, reducing function call overhead and improving cache utilization. This approach is used in systems like Apache Arrow, DuckDB, and modern Spark SQL execution engines. It significantly improves performance for analytical workloads by leveraging SIMD (Single Instruction Multiple Data) CPU instructions.