中文摘要：

云计算的一个关键需求是其基础设施中大规模虚拟机的放置问题.虚拟机和物理结点之间的映射决定了如何将云计算中虚拟化资源分配给多个Web应用,对云计算系统的性能、能耗和QoS保证有重要影响.文中提出了云计算中虚拟机放置的自适应管理框架,提出了带应用服务级目标约束的虚拟机放置多目标优化遗传算法,用于制定框架中的虚拟机放置策略.算法基于长期负载性能模型,采用组方式和三空间分割方法分别对染色体进行编码和译码,根据不同染色体长度的变化设计交叉和变异遗传算子.算法对解空间内的多个区域同时搜索,具有群体和自我进化的优势,优化一次就能获得对不同目标的权值运算多次才能得到的最优解.实验结果表明,与传统的启发式和单目标优化算法相比,提出的框架及算法使得多个应用的服务级目标的违背率最低,且能有效减少虚拟机迁移次数和物理结点的使用数量.

英文摘要：

Virtual machine placement in the cloud infrastructure is an important problem that remains to be effectively addressed.The mapping problem between virtual machines and physical nodes is to decide how to allocate virtualized resources on the cloud to many Web applications,thus it greatly impacts on the performance,cost and QoS guaranteed service.An adaptive management framework for virtual machine placement in the cloud is proposed.A multi-objective optimization genetic algorithm is presented to determine placement strategy in the framework,which is subject to application service level objects（SLOs） constraint.It encodes the chromosome using the group method,and crossover and mutation operations deal with the chromosome which length is varying.It decodes the chromosome using three-dimensional split method.The experimental results show that,the proposed solution could effectively reduce the number of used nodes and virtual machine migration,and minimize violation of many application SLOs,compared with traditional heuristic methods and single objective solution.