Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

Competitive Analysis of Distributed Algorithms James Aspnes?

Summary: Competitive Analysis of Distributed Algorithms
James Aspnes?
Yale University, Department of Computer Science
Abstract. Most applications of competitive analysis have involved on-
line problems where a candidate on-line algorithm must compete on
some input sequence against an optimal o -line algorithm that can in
e ect predict future inputs. E orts to apply competitive analysis to fault-
tolerant distributed algorithms require accounting for not only this in-
put nondeterminism but also system nondeterminism that arises in dis-
tributed systems prone to asynchrony and failures. This paper surveys
recent e orts to adapt competitive analysis to distributed systems, and
suggests how these adaptations might in turn be useful in analyzing a
wider variety of systems. These include tools for building competitive al-
gorithms by composition, and for obtaining more meaningful competitive
ratios by limiting the knowledge of the o -line algorithm.
1 Introduction.
Like on-line algorithms, distributed algorithms must deal with limited informa-
tion and unpredictable user and system behavior. Unlike on-line algorithms, in
many distributed algorithms the primary source of di culty is the possibility
that components of the underlying system may fail or behave badly. In a dis-


Source: Aspnes, James - Department of Computer Science, Yale University


Collections: Computer Technologies and Information Sciences