The anonymous consensus hierarchy and naming problems

Abstract

This paper investigates whether the assumption of unique identifiers is essential for wait-free distributed computing using shared objects of various types. Algorithms where all processes are programmed identically and do not use unique identifiers are called anonymous. We study the anonymous solvability of two key problems, consensus and naming. These problems are used to define measures of a type T's power to solve problems anonymously. These measures provide a significant amount of information about whether anonymous implementations of one type from another are possible. We compare these measures with one another and with the consensus numbers defined by Herlihy [13]. © Springer-Verlag Berlin Heidelberg 2007.