中文摘要：

将服务部署优化问题建模为多目标组合优化问题.在多目标遗传算法的基础上,把部署方案转换为基因编码,用轮盘赌选择机制选择个体,用单点交叉算子产生新的子代,并以设定的概率发生变异.对合适个体考虑支配值和稀疏值设计适应度函数;对不合适个体根据支配值和SLA冲突设计适应度函数.最后给出了优化过程.通过仿真实验可以看出:随着迭代次数的增加,适应度值及各个优化指标值逐渐收敛于一个固定且较优值,说明利用设计的优化算法,能使各个优化目标值较快地收敛到一个较优解,能较好地帮助基础设施即服务(SaaS)提供商在部署应用服务时进行有效规划和决策.

英文摘要：

The optimization of service deployment was modeled as multi-objective composition optimization. On the basis of multi-objective genetic algorithm, deployment solutions were converted to gene codes; individuals were selected by roulette mechanism; new generations were produced by single point crossover operator; variations were appeared in preset probability. The fitness function for fit individuals was based on dominant value and sparse value, while this function for unfit individuals was based on dominant value and SLA collisions. The optimization process was also proposed. In simulation experiment the fitness value and other evaluations converge at a fixed and general optimal value gradually when iteration number increases. It shows that the optimization method in this paper can help multi-objective values converge at a general optimal value and give help to infrastructure as a service (SaaS) provider on planning and decision of service deployment. ? 2016, Editorial Board of Journal of Huazhong University of Science and Technology. All right reserved.