Skip trees, an alternative data structure to skip-lists in a concurrent approach

Abstract

We present a new type of search trees isomorphic to Skip-lists, i.e., there is a one-to-one mapping that commutes with the update algorithms. Moreover, as Skip-trees inherit all the race properties from Skip-lists, they can be selected as well as Skip-lists. We design a concurrent algorithm on the fly approach to update skip-trees. Among other advantages, this algorithm is more compressive than the one designed by Pugh for Skip-lists. It is based on the closest relationship between Skip-trees and the family of B-trees that allows the translation of local rules between them. From a practical point of view, although the Skip-list should be in main memory, the Skip-trees can be registered into secondary external storage, therefore it can be likely to analyse its ability to manage Databases.