Distributed Computing

, Volume 8, Issue 3, pp 115–120 | Cite as

On the complexity of global computation in the presence of link failures: the general case

  • Y. Afek
  • D. Hendler
Article

Summary

This paper presents Ω(m logn) and Ω(mn) messages lower bounds on the problem of computing a gobal sensitive function in biderectional networks with link failures (i.e., dynamically changing topology), wheren andm are the total number of nodes and links in the network. The Ω(m logn) lower bound is under the assumption thatn is a-priori known to the nodes, while the second bound is for the case in which such knowledge is not available. A global sensitive function ofn variables is a function that may not be computed without the knowledge of the values of all then variables (e.g. maximum, sum, etc). Thus, computing such a function at one node of a distributed network requires this node to communicate with every other node in the network. Though lower bounds higher than Ω(m) messages are known for this problem in the context of link failures, none holds for dense bidirectional networks. Moreover, we are not aware of any other nontrivial lower bound higher than Ω(m) for dense bidirectional networks.

Key words

Lower bounds Distributed computing 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Frederickson GN, Lynch NA: Electing a leader in a synchronous ring. ACM 34(1): 98–115 (1987)Google Scholar
  2. 2.
    Pach J, Korach E, Rotem D: A technique for proving lower bounds for distributed maximum-finding algorithms. In: Proc of the 14th Annu ACM Symp on Theory of Computing, pp 378–382, 1982Google Scholar
  3. 3.
    Frederickson GN, Lynch NA: A general lower bound for electing a leader in a ring. Tech Rep CSD-TR-512. Purdue University, 1985Google Scholar
  4. 4.
    Goldreich O, Shrira L: The effect of link failures on computations in asynchronous rings. In: Proc of the ACM Symp on Principles of Distributed Computing, 1986Google Scholar
  5. 5.
    Linial N: Distributive graph algorithms — global solutions from local data. In: Proc of the 28th IEEE Annu Symp on Foundation of Computer Science, pp 331–335, 1987Google Scholar
  6. 6.
    Cidon I, Shavitt Y: Message terminate algorithms for rings of unknown size. In: Proc of the 6th Int Workshop on Distributed Algorithms. Lect Notes Comput Sci, vol 647. Springer, Berlin Heidelberg New York 1992, pp 264–276Google Scholar
  7. 7.
    Goldreich O, Shrira L: Consultation in the presence of faults: two lower bounds. Tech Rep, Technion, 1985Google Scholar
  8. 8.
    Afek Y, Landau GM, Schieber B, Yung M: The power of multimedia: combining point-to-point and multiaccess networks. Inf Comput 84(1): 97–118 (1990).Google Scholar
  9. 9.
    Segall A: Distributed network protocols. IEEE Trans Inf Theory, IT-29(1): 23–35 (1983)Google Scholar
  10. 10.
    Goldreich O, Shrira L: On the complexity of computation in the presence of link failures: the case of a ring. Distrib Comput 5: 121–131 (1991)Google Scholar
  11. 11.
    Goldreich O, Sneh D: On the complexity of global computation in the presence of link failures: the case of uni-directional faults. In: Proc 11 th ACM Symp on Principles of Distributed Computing, pp 103–111, 1992Google Scholar
  12. 12.
    Afek Y, Gafni E: End-to-end communication in unreliable networks. In: Proc of the 7th ACM Symp on Principles of Distributed Computing, pp 131–148, 1988Google Scholar
  13. 13.
    Afek Y, Gafni E: Bootstrap network resynchronization: an efficient technique end-to-end communication. In: Proc of the 10th Annu ACM Symp on Principles of Distributed Computing (PODC), 1991Google Scholar

Copyright information

© Springer-Verlag 1995

Authors and Affiliations

  • Y. Afek
    • 1
  • D. Hendler
    • 1
  1. 1.Computer Science DepartmentTel-Aviv UniversityRamat AvivIsrael

Personalised recommendations