On the computational power of self-stabilizing systems

James Abello, Shlomi Dolev

Research output: Contribution to journalArticlepeer-review

5 Scopus citations

Abstract

The computational power of self-stabilizing distributed systems is examined. Assuming availability of any number of processors, each with (small) constant size memory we show that any computable problem can be realized in a self-stabilizing fashion. The result is derived by presenting a distributed system which tolerates transient faults and simulates the execution of a Turing machine. The total amount of memory required by the distributed system is equal to the memory used by the Turing machine (up to a constant factor).

Original languageEnglish
Pages (from-to)159-170
Number of pages12
JournalTheoretical Computer Science
Volume182
Issue number1-2
DOIs
StatePublished - 15 Aug 1997

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'On the computational power of self-stabilizing systems'. Together they form a unique fingerprint.

Cite this