Sorting And Indexing With Partitioned B-Trees

Partitioning within a B-tree, based on an artificial leading key column and combined with online reorganization, can be exploited during external merge sort for accurate deep read-ahead and dynamic resource allocation, during index creation for a reduced delay until the first query can search the new index, during data loading for streaming integration of new data into a fully indexed database, and for miscellaneous other operations. Despite improving multiple fundamental database operations using a single basic mechanism, the proposal offers these benefits without requiring data structures or algorithms not yet supported in modern relational database management systems. While some of the ideas discussed here have been touched upon elsewhere, the focus here is on re-thinking the relationship between sorting and B-trees more thoroughly, on exploiting this relationship to simplify and unify data structures and algorithms, and on gathering comprehensive lists of issues and benefits.

• Type: Conference
• Year: 2003
• Where: CIDR
• Authors: Goetz Graefe
• Keywords: CIDR 2003, CIDR 2007, Data Structures, External Merge Sort, Relational Database Management Systems