TC3PoP: a time-cost compromised workflow scheduling heuristic customized for cloud environments

Abstract

Recently, optimizing the monetary cost and performance metrics of executing workflow applications in cloud environments has become an important and interesting research subject. The most critical challenges with the existing methods are simplifying assumptions, which make them far from reality, offering a set of solutions that cannot cover the whole search space and are not real non-dominated solutions, and finally, their high time-complexity. To tackle the mentioned problems, in this paper, a multi-objective workflow scheduling algorithm named Time-Cost Compromised CPoP (TC3PoP) customized for cloud environments is proposed. We have modeled and formulated the workflow scheduling problem based on the cloud environment offers and characteristics. The proposed algorithm is light-weight and based on the offered resources by cloud providers, and produces a set of real Pareto optimal solutions so that a user can freely choose the best solution based on his/her budget and deadline. The results of the experiments show that in terms of HyperVolume indicator, the validity and diversity of TC3PoP solutions are 41.5% and 38.1% better than those of well-known and successful NSGA-II for two real scientific workflows, Montage and LIGO, respectively, and the proposed method is much faster than the compared method. Moreover, to evaluate the quality of Pareto fronts produced by the proposed algorithm, some statistical analysis was made to investigate the precision and distribution of the solutions.