How Index and B-Tree Index Work
Introduction Index When creating an Index, PostgreSQL creates a separate physical file on the disk. Each index has a Relfilenode which is a unique identifier number. An index does not store the entire information of a row but only contains Index Entries, each entry including: Key: The value of the column you index. TID (Tuple Identifier): A physical pointer consisting of a BlockNumber and an OffsetNumber, used for reference to point to the location of that row in the main table (HEAP). When executing a query, the Query Planner Cost Model in Postgres will calculate to choose between reading data from the index or retrieving it directly from the HEAP (Sequential Scan). {Index Scan Cost} = {Index file reading cost} + {Random block reading cost in the table} {Seq Scan Cost} = {Sequential block reading cost in the table} Because the Index only contains the TID to point to the data in the main table, after loading the index content and finding the necessary values, it must perform random I...