Predicate Pushdown

Predicate pushdown applies WHERE conditions at the earliest possible point: if a query selects customers where age > 30, pushing the predicate down means applying the filter at the table scan (only read rows matching age > 30) rather than reading all customers then filtering. This dramatically reduces data movement: a 1 billion row customer table filtered to 100 million matching rows avoids moving 900 million unnecessary rows. Pushdown is especially valuable in distributed systems: filtering at the source node reduces network traffic between nodes and reduces memory requirements.

Predicate pushdown is particularly powerful in columnar systems: filters on one column avoid reading other columns entirely. A query filtering on age but selecting name/email reads only age column from storage, never touches name/email columns. Modern query optimizers automatically implement pushdown: users write intuitive WHERE conditions, optimizers push them down automatically.

In practice, predicate pushdown occurs at multiple levels: source database applies filters before sending data, columnar stores apply column filters during scan, distributed systems apply filters at partition level. Effective optimization requires indexes and statistics: if optimizer doesn't know that age > 30 filters to 10% of rows, it may not choose to filter early.