Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

Wait-free Byzantine Consensus College of Computer Science, Northeastern University and

Summary: Wait-free Byzantine Consensus
Paul Attie
College of Computer Science, Northeastern University and
MIT Laboratory for Computer Science
Keywords: distributed computing, fault tolerance, wait-freedom
1 Introduction
Byzantine consensus has traditionally been studied in the context of message passing models of dis-
tributed computing, with various asynchrony/timing assumptions on message transit times and/or
process speeds [2, 4]. In an asynchronous shared memory model, Byzantine consensus has been
considered \uninteresting" because a single Byzantine process can repeatedly corrupt the entire
memory, thereby overwhelming any consensus algorithm. One way to address this problem is to
impose assumptions on the relative speeds of processes, even Byzantine ones, so that a process can
do only a limited amount of damage within a xed time interval. Another possibility is to limit the
amount of memory that each process can access, e.g., through operating system level mechanisms
such as access control matrices, access control lists, or capability lists [8, 10, 11]. We consider the
latter approach in this paper.
A consensus protocol must have some degree of fault-tolerance (otherwise the problem is trivial).
Among the attributes of a fault are: (1) the entity su ering the fault, e.g., the memory itself, or a
process that accesses the memory, and (2) the type of the fault, e.g., crash (entity stops responding),
omission (entity responds to only some inputs), or Byzantine (entity behaves arbitrarily). We


Source: Attie, Paul - Department of Computer Science, American University of Beirut


Collections: Computer Technologies and Information Sciences