Query Execution Engines: Volcano Iterator Model, Vectorized Batch Processing (MonetDB/X100, DuckDB, Velox), and HyPer-Style LLVM JIT Query Compilation

Deep End Era. Engineering with Sankalp · Oct 29, 2026 · 1h 9m

A hundred-million-row smart-meter aggregation query is running entirely in RAM on a modern superscalar processor—zero disk reads, zero lock waits—and still takes fourteen seconds while leaving ninety percent of the CPU's execution units idle. When the same query runs on a vectorized engine or an LLVM JIT-compiled engine, it finishes in one hundred and eighty milliseconds. To understand where the other thirteen point eight seconds went, we dissect the Volcano iterator model, MonetDB/X100 vectorization in DuckDB and Velox, and Thomas Neumann's HyPer query compilation.

Chapters

00:00:00 Cold…