Scheduling with Complete Multipartite Incompatibility Graph on Parallel Machines

TL;DR

This paper investigates scheduling on parallel machines with incompatibilities modeled as a complete multipartite graph, providing algorithms for optimal and approximate solutions across various machine types and criteria, including new LP-based methods.

Contribution

The paper introduces new algorithms and complexity results for scheduling with multipartite incompatibilities, including a novel LP relaxation approach for total completion time on uniform machines.

Findings

Delimits polynomial and NP-hard cases for various machine types.

Provides approximation algorithms with guaranteed ratios and PTAS for hard cases.

Introduces a novel LP relaxation technique for total completion time on uniform machines.

Abstract

In this paper we consider the problem of scheduling on parallel machines with a presence of incompatibilities between jobs. The incompatibility relation can be modeled as a complete multipartite graph in which each edge denotes a pair of jobs that cannot be scheduled on the same machine. Our research stems from the work of Bodlaender et al.~[1992, 1993]. In particular, we pursue the line investigated partially by Mallek et al.~[2019], where the graph is complete multipartite so each machine can do jobs only from one partition. We also tie our results to the recent approach for so-called identical machines with class constraints by Jansen et al.~[2019], providing a link between our case and their generalization. In the paper we provide several algorithms constructing schedules, optimal or approximate with respect to the two most popular criteria of optimality: Cmax (the makespan) and {\Sigma}Cj(the total completion time). We consider a variety of machine types in our paper: identical, uniform, unrelated, and a natural subcase of unrelated machines. Our results consist of delimitation of the easy (polynomial) and NP-hard problems within these constraints. In the case when the problem is hard, we also provide algorithm, either with a guaranteed constant worst-case approximation ratio or even in some cases a PTAS. In particular, we fill the gap on research for the problem of finding a schedule with smallest total completion time on uniform machines. We address this problem by developing a linear programming relaxation technique with an appropriate rounding, which to our knowledge is a novelty for this criterion in the considered setting.