Deterministic monotone algorithms for scheduling on related machines

  • Authors:
  • Pasquale Ambrosio;Vincenzo Auletta

  • Affiliations:
  • Dipartimento di Informatica ed Applicazioni R.M. Capocelli, Universitá di Salerno, via Ponte Don Melillo, I-84084 Fisciano (SA), Italy;Dipartimento di Informatica ed Applicazioni R.M. Capocelli, Universitá di Salerno, via Ponte Don Melillo, I-84084 Fisciano (SA), Italy

  • Venue:
  • Theoretical Computer Science
  • Year:
  • 2008

Quantified Score

Hi-index 5.23

Visualization

Abstract

We consider the problem of designing monotone deterministic algorithms for scheduling tasks on related machines in order to minimize the makespan. Several recent papers showed that monotonicity is a fundamental property to design truthful mechanisms for this scheduling problem. We give both theoretical and experimental results. First of all we consider the case of two machines when speeds of the machines are restricted to be powers of a given constant c0. We prove that algorithm Largest Processing Time (LPT) is monotone for any c=2 while it is not monotone for c@?1.78; algorithm List Scheduling (LS), instead, is monotone only for c2. In the case of m2 machines we restrict our attention to the class of ''greedy-like'' monotone algorithms defined in [Vincenzo Auletta, Roberto De Prisco, Paolo Penna, Giuseppe Persiano, Deterministic truthful approximation mechanisms for scheduling related machines, in: Proceedings of 21st Annual Symposium on Theoretical Aspects of Computer Science. STACS '04, in: Lecture Notes in Computer Science, vol. 2996, Springer, 2004, pp. 608-619]. It has been shown that greedy-like monotone algorithms can be used to design a family of 2+@e-approximate truthful mechanisms. In particular, in [Vincenzo Auletta, Roberto De Prisco, Paolo Penna, Giuseppe Persiano, Deterministic truthful approximation mechanisms for scheduling related machines, in: Proceedings of 21st Annual Symposium on Theoretical Aspects of Computer Science. STACS '04, in: Lecture Notes in Computer Science, vol. 2996, Springer, 2004, pp. 608-619], the greedy-like algorithm Uniform is proposed and it is proved that it is monotone when machine speeds are powers of a given integer constant c0. In this paper we propose a new algorithm, called Uniform_RR, that is still monotone when speeds are powers of a given integer constant c0 and we prove that its approximation factor is not worse than that of Uniform. We also experimentally compare the performance of Uniform, Uniform_RR, LPT, and several other monotone and greedy-like heuristics.