The Random Buffer Tree : A Randomized Technique for I/O-efficient Algorithms

Abstract

In this paper, we present a probabilistic self-balancing dictionary data structure for massive data sets, and prove expected amortized I/O-optimal bounds on the dictionary operations. We show how to use the structure as an I/O-optimal priority queue. The data structure, which we call as the random buffer tree, abstracts the properties of the random treap and the buffer tree and has the same expected I/O-bounds as the buffer tree.