The query execution engine plays an important role in database system performance. TiDB, an open-source MySQL-compatible Hybrid Transactional/Analytical Processing (HTAP) database, implemented the widely-used Volcano model to evaluate queries. Unfortunately, when querying a large dataset, the Volcano model caused high interpretation overhead and low CPU cache hit rates.
Inspired by the paper MonetDB/X100: Hyper-Pipelining Query Execution, we began to employ vectorized execution in TiDB to improve query performance. (Besides this article, we also suggest you take Andy Pavlo’s course on Query Execution, which details principles about execution models and expression evaluation.)