Faster hash joiner with vectorized execution

Faster hash joiner with vectorized execution

By contrast, in the vectorized execution model, each component of the plan processes an entire batch of columnar data at once, instead of just a single row. The vectorized execution model aims at this simple truth by replacing the fully-general, interpreter-like SQL expression evaluator with a series of specific compiled loops, specialized on datatype and operation, so that the computer can do many more simple tasks in a row before having to step back and make a decision about what to do next. In the Volcano model, each data row is processed one at a time by each operator – a row-by-row execution approach.

Source: www.cockroachlabs.com