A compact index structure with high data retrieval efficiency

Abstract

The B <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">+</sup> -tree and its variants have been reported as the good index structures for retrieving data. However, their index schemes are short of optimizing data arrangement in the external structures. Improper handles adopted in the traditional splitting policies including pre-partition of data underlying space, node splitting by force, node splitting with unbalanced partition, and node splitting upon overflowed loading usually burden index structures with plenty of inefficient storage space and exceeding construction overhead. Consequentially, time and space efficiencies in conventional B <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">+</sup> -tree and its variants remain much room for improvement. In this paper, a new index scheme is proposed to organize data into a more compactness via a better splitting policy. The new index scheme constantly compresses data in the external structure of an index tree and makes neighboring data more aggregative than traditional methods. The compact B <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">+</sup> -trees elevate space utilization to a higher level which is near 90%. Both system efficiencies and data selectivity are remarkably improved by compact B <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">+</sup> -trees.