Teradata's MPP architecture requires an efficient way to distribute and retrieve data, and does so using hash partitioning. Most tables in Teradata use hashing to distribute rows across AMPs based on the Primary Index (PI) value for each row. The data is stored in the Block File System (BFS), and Teradata may scan the entire table or use indexes to access the data, depending on the query. This approach ensures scalable performance and efficient data access.
- If the Primary Index is unique, the rows in the tables are automatically distributed evenly by hash partitioning.
- The designated Primary Index column or columns are hashed to generate consistent hash codes for the same values.
- No reorganization, repartitioning, or space management is required.
- Each AMP typically contains rows from all tables, ensuring efficient data access and processing.