Adaptive Index Structures

Traditional indexes aim at optimizing the node accesses during query processing, which, however, does not necessarily minimize the total cost due to the possibly large number of random accesses. In this paper, we propose a general framework for adaptive indexes that improve overall query cost. The performance gain is achieved by allowing index nodes to contain a variable number of disk pages. Update algorithms dynamically re-structure adaptive indexes depending on the data and query characteristics. Extensive experiments show that adaptive B- and R-trees significantly outperform their conventional counterparts, while incurring minimal update overhead.

• Type: Journal
• Year: 2002
• Where: VLDB
• Authors: Yufei Tao, Dimitris Papadias
• Keywords: Adaptive Indexes, Database, Query Processing, Traditional Indexes, VLDB 2002