Certification trails for data structures

Abstract

The applicability of the certification trail technique, a recently introduced and promising approach to fault detection and fault tolerance, is expanded. Previously, certification trails had to be customized to each algorithm application, but here trails appropriate to wide classes of algorithms are developed. These certification trails are based on common data-structure operations such as those carried out using balanced binary trees and heaps. Any algorithm using these sets of operations can therefore employ the certification trail method to achieve software fault tolerance. Constructions of trails for abstract data types such as priority queues and union-find structures are given. These trails are applicable to any data structure implementation of the abstract data type. It is shown that these ideas lead naturally to monitors for data structure operations.<>