7 - Range partitioning

Abstract

This chapter discusses essential concept of range partitioning. Range partitioning is one of the surprising techniques of physical database design. The essential concept of range partitioning is to divide a single table into multiple little tables, so that each little table holds a specific range of the data. From an application perspective the collection of partitions is accessed as a single table, and the application need not be programmed to be aware of the partitioning design. Range partitioning is used both to improve the performance and ease of administrative tasks like roll-in and roll-out and to improve query execution plan selection for decision support queries by means of partition elimination. It divides a large table into several smaller ones. Even though the table remains a single table from the application perspective, the database kernel will store each partition in its own table object. As a result, each partition has the supported capacity of a single table. The major factor limiting the size of a table is usually the addressability of the records within the tables, which is an attribute of the record identifier (RID) architecture. Some RID architectures support a huge addressability domain that removes the need to exploit range partitioning for purely increased table size.