Abstract: Data placement for industrial workflows in hybrid cloud environments presents significant challenges,involving the assurance of data security and the balancing of interests between users and service providers,while taking into consideration factors including data transfer latency,workflow execution cost,and load balancing among data centers.To address these challenges,an Improved Optimization Multi-objective Optimization Evolutionary Algorithm (IO-MOEA) based data placement strategy was proposed.This approach improved its convergence and diversity by adaptively enhancing the selection operator within the fast elitist Non-dominated Sorting Genetic Algorithm(NSGA-II).Furthermore,the entropy weight method and the Technique for Ordering Preference by Similarity to Ideal Solution (TOPSIS) method were combined to objectively evaluate the advantages and disadvantages of the solutions in the Pareto optimal set to find the best one.Experimental results showed that the proposed algorithm could effectively reduce the data transfer time and execution cost for industrial workflows,while ensuring the load balancing among data centers.Compared to the original algorithm,the IO-MOEA algorithm improved the hypervolume by about 3% ~19% and the space by about 11%~21%.

Key words: cloud computing, industrial workflow, multi-objective optimization, data placement, load balancing

摘要： 针对混合云环境下工业软件工作流的数据放置问题,如何在保证数据安全的前提下平衡用户和服务提供商的利益,综合考虑数据的传输时延,工业软件工作流执行代价以及数据中心间的负载是一个重要的挑战。为此,提出一种安全等级分级机制,并设计出一种基于改进的多目标优化进化算法(IO-MOEA)的数据放置策略。该策略在传统非支配排序遗传算法(NSGA-II)中对选择算子进行自适应改进,提高了算法的收敛性和种群的多样性,之后结合熵权法和理想解相似性排序偏好技术(TOPSIS)法,客观评估Pareto最优解集中解的优劣,从而找到最佳方案。实验结果表明,所提算法能够有效降低工业软件工作流传输时间和执行代价,同时兼顾数据中心间的负载均衡。相比于改进前的算法,改进后的IO-MOEA算法在超平面指标上提高了约3%~19%,在空间指标上改善了11%~21%。

关键词: 云计算, 工业软件工作流, 多目标优化, 数据放置, 负载均衡