Talel Ladhari

Deployment of Solving Permutation Flow Shop Scheduling Problem on the Grid

This paper describes a parallel algorithm solving the m-machines, n-jobs, permutation flow shop scheduling problem as well as its deployment on a Grid of computers (Grid’5000). Our algorithm is basically a parallelization of the well known Branch and Bound method, which is an exact method for solving combinatorial optimization problems. We present, in particular, a new strategy of parallelization which uses some directives of communication between all processors in order to update the value of the Upper Bound. The purpose is to obtain the optimal scheduling of all the jobs as quickly as possible by modifying and adapting the classical sequential Branch and Bound method. The deployment of our application on the Grid gives good results and permits the resolution of new instances not yet resolved.

A Load Balanced Distributed Algorithm to Solve the Permutation Flow Shop Problem Using the Grid

This paper deals with solving the permutation flow shop problem (PFSP) which is an NP-hard scheduling problem using grid computing. As the sequential resolution of this problem can be sometimes impossible on a single machine and especially for instances of large data, we are interested in this work by solving the PFSP using parallel programming. In a previous paper, we presented a parallel algorithm for solving the PFSP on the grid called GAUUB. The GAUUB algorithm uses the Branch and Bound method to find optimal solutions of the problem and distributes the tasks among all processors. In this paper, we present a new algorithm called GALB, based on a parallelization strategy ensuring better load balancing between the processors and therefore, better efficiency of the algorithm. Computational results of our new algorithm performed on the grid showed good results and significant improvements of our initial algorithm thanks to our load balancing technique.

Strong multi-commodity flow formulations for the asymmetric traveling salesman problem

Abstract We provide new compact formulations of polynomial size for the asymmetric traveling salesman problem obtained through the Reformulation-Linearization Technique. The first one is obtained directly by this latter approach while the two others are derived by performing projections of this formulation on the variables of the existing models. We show that the devised formulations are stronger than the state-of-the-art models. Computational experiments conducted on benchmark instances for the classical variant and with precedence constraints confirm the better quality of the relaxations provided by our proposed formulations.

Nodes Migration Scheduling of Access Networks

In this paper, we consider the nodes migration scheduling problem that aims to migrate nodes from an outdated access network to a new one. In order to avoid services’ interruption, a bridge is installed temporarily between the two networks and, as for the nodes, they are migrated, one at each time period, from the old to the new network. The objective is to find a migration’s sequence that minimizes the cost of the capacity to install on the bridge. We provide a compact formulation of polynomial size in order to solve the problem optimally. We reformulate it in such a way that the new model reduces drastically the branch-and-bound search tree. In addition, we give a lower bound based on a bi-partitioning problem that turns out to be close to the optimal solutions. We conduct experiments on a case study consisting of migrating eNodeBs of a 4G network from an access network to another. The computational experiments show the performance of our approach as the proposed model provided optimal solutions for up to 40 nodes.

An Efficient Iterated Greedy Algorithm for the Makespan Blocking Flow Shop Scheduling Problem

We propose in this paper a Blocking Iterated Greedy algorithm (BIG) which makes an adjustment between two relevant destruction and construction stages to solve the blocking flow shop scheduling problem and minimize the maximum completion time (makespan). The greedy algorithm starts from an initial solution generated based on some well-known heuristic. Then, solutions are enhanced till some stopping condition and through the above mentioned stages. The effectiveness and efficiency of the proposed technique are deduced from all the experimental results obtained on both small randomly generated instances and on Taillards benchmark in comparison with state-of-the-art methods.

Genetic local search algorithm for minimizing the total completion time in single machine scheduling problem with release dates and precedence constraints

In this work we investigate the single machine scheduling problem with release dates and precedence constraints for minimizing the sum of completion times. To solve this problem approximately, we propose three constructive heuristics as well as a genetic local search algorithm. Computational results show that the proposed genetic local search algorithm yields near-optimal solutions and provides interesting results.