Schedule length and reliability-oriented multi-objective scheduling for distributed computing

Maximizing system reliability and minimizing schedule length are the two major objectives in scheduling a distributed computing system. These two objectives have been considered separately by most researchers, although more realistically they should be considered simultaneously. This paper addresses the problem by taking a multi-objective approach in scheduling. A Tabu search algorithm is proposed and two lateral interference schemes are used to distribute the Pareto optimal solutions along the Pareto front uniformly. Randomly generated directed acyclic graphs and a real application task graph are used to study the performance of the proposed algorithms. Experimental results show that for this problem lateral interference has no influence on the non-dominated solution number, but does benefit the uniform distribution of non-dominated solutions, irrespective of the computation method used to determine distances between the solutions.