Mixed satisfiability tests for multiprocessor scheduling with release dates and deadlines
Introduction
In the identical parallel machine problem, a set I of activities has to be processed on m identical parallel machines. Each activity, Ai∈I, has a release date ri, a processing time pi and a deadline . All the data are assumed to be deterministic and integer. Each machine processes at most one job at a time and each job cannot be processed by more than one machine. We also assume that preemption is not allowed, and that machines are available from time zero onwards. Our aim is either to find a schedule in which activities are processed within their time-window , without violating the machine constraint, or to prove that there exists no such solution. Classically, this problem is denoted by , which is known to be -hard in the strong sense. In this paper, we address some techniques to detect that there exists no feasible solution.
Our main motivation for studying this problem is the fact that it can be seen as a relaxation of some fundamental complex scheduling problems as Hybrid Flowshop and RCPSP. It corresponds to the decisional variant of P|ri|Lmax [6] and then it is strongly connected to the P|ri,qi|Cmax, for which several efficient lower bounds have been presented, e.g., [4], [5], [6].
In this paper, we focus on two satisfiability tests for : the first one is based on the preemptive relaxation, that can be solved in polynomial time, the second one is the energetic reasoning. Our basic motivation is to use the complementarity of those two approaches, using the notion of mandatory parts of activities into the classical max-flow formulation.
In Section 2, both preemptive approach (PR) and energetic reasoning (ER) applied to the are recalled and examples illustrating the drawbacks of those approaches are presented. In Section 3, we propose a new model that is based on a max-flow with minimum capacity. In Section 4, we show that our approach dominates recent satisfiability tests the presentation of which does not mention energetic reasoning, even if they implicitly use it.
Section snippets
Classical satisfiability tests for
Satisfiability tests provide some necessary conditions for the existence of a feasible solution. These tests play a central part for both computing destructive lower bounds and finding optimal solutions to -hard problems such as the Resource Constrained Project Scheduling Problem [2]. We briefly recall the two classical approaches that are used to get satisfiability tests for the .
Mixing max-flow approach and energetic reasoning
This section is devoted to an improvement of the satisfiability test based on PR. It also identifies a set of activities that have a non-zero mandatory part over a time interval. These mandatory parts are integrated in graph G of PR as minimum constraints on some edges. The mandatory part of each activity is given using ER. First, we introduce the new set of time points to add to the initial formulation of PR. Proposition 1 Let Ai∈I be an activity such that , then W(i,t1,t2)=t
Comparison with Haouari and Gharbi's approach
Proposition 3 The max-flow formulation using mandatory parts of activities is better than the one presented in [6].
In a recent paper, Houari and Gharbi [6] have proposed to build semi-preemptive schedule. It is based on a technical result to enforce some parts of an activity to be processed into a given time interval. They proved that if an activity Ai verifies , then one machine processes activity Ai during .
From the energetic reasoning point of view, it is obvious that
Acknowledgements
The authors are grateful to all members of the French research group GOTHA-Constraints and Scheduling for their helpful comments.
References (7)
- et al.
An improved max-flow-based lower bound for minimizing maximum lateness on identical parallel machines
Oper. Res. Lett.
(2003) - et al.
Network flows: Theory, Algorithms and Applications
(1993) - et al.
Satisfiability tests and time-bound adjustments for cumulative scheduling problems
Ann. Oper. Res.
(1999)