基于Seq2Seq模型的工作流动态调度多目标进化算法

Multi-objective Evolutionary Algorithm for Workflow Dynamic Scheduling Based on Seq2Seq Model

将数据处理类工作流在云计算环境下的调度问题建模为动态多目标优化问题,同时为了解决静态多目标优化算法在环境参数动态变化下可能出现的种群多样性缺失问题,在NSGA-II算法的基础上结合Seq2Seq深度学习模型,提出了DNSGA-II-Seq2Seq算法,算法通过Seq2Seq模型学习连续历史环境下局部最优解的变化规律,在环境变化时预测新的解并将其加入NSGA-II算法的种群中,以解决多样性缺失问题,同时加速算法收敛。在改进的WorkflowSim上进行的实验表明,与其他经典的算法相比,DNSGA-II-Seq2Seq算法预测的解和最终结果在多项指标上均优于其他算法,验证了算法的有效性。

The scheduling problem of data processing workflow in cloud computing environment was modeled as a dynamic multi-objective optimization problem. At the same time, in order to solve the possible lack of population diversity of static multi-objective optimization algorithm with the dynamic change of environmental parameters, DNSGA-II-Seq2 Seq algorithm was proposed based on NSGA-II algorithm and Seq2 Seq deep learning model. The algorithm could learn the change law of the local optimal solution in the continuous historical environment through the Seq2 Seq model, predict the new solution when the environment change, and adds it to the population of NSGA-II algorithm to solve the problem of lack of diversity and accelerate the convergence of the algorithm. Experiments on the improved WorkflowSim showed that compared with other classical algorithms, the predicted solution and final result of DNSGA-II-Seq2 Seq algorithm were better than other algorithms in many indexes, which verified the effectiveness of the algorithm.