绿色数据中心流量感知虚拟机迁移(英文)

绿色数据中心内将虚拟机迁移并整合到更少数目的物理机上,从而关闭或睡眠空闲的物理机,是非常高效的数据中心节能方法.本文构建了新的流量感知虚拟机迁移模型,把物理机使用数量、流量代价、迁移代价影响都置入模型,使得虚拟机的迁移方案达到缩减物理机目的的同时,减少网络代价和迁移过程中的代价。提出了最小流量代价优先匹配的贪婪算法来解决这个多目标优化问题。仿真结果显示,这个多目标的优化中,多个优化目标之间存在一定的冲突,因此需要折中处理。该算法在物理机数量和迁移代价两个目标上做出了微小的让步,得到了较好的流量代价优化。实际的迁移算法应用中,还应考虑各个部分的真实能耗,采取合理的折中处理.     

Optimization of loss and gain multilayers for reducing the scattering of a perfect conducting cylinder

Reduction of electromagnetic scattering from a conducting cylinder could be achieved by covering it with optimized multilayers of normal dielectric and plasmonic material. The plasmonic material with intrinsic losses could degrade the cloaking effect. Using a genetic algorithm, we present the optimized design of loss and gain multilayers for reduction of the scattering from a perfect conducting cylinder. This multilayered structure is theoretically and numerically analyzed when the plasmonic material with low loss and high loss respectively is considered. We demonstrate by full-wave simulation that the optimized nonmagnetic gain-loss design can greatly compensate the decreased cloaking effect caused by loss material,which facilitates the realization of practical electromagnetic cloaking, especially in the optical range.     

最小化成本的可靠虚拟基础设施设计(英文)

在保证虚拟网络可靠性需求的条件下以最小化映射成本为目标,研究了可靠虚拟基础设施映射.在RVIM算法的基础上提出了Improved_RVIM算法,算法中一些工作虚拟节点适当地映射在相同的设施节点上,可以在不增加备份计算资源的条件下节省通信成本,降低总映射成本.仿真结果表明Improved-RVIM算法相比其它算法成本更低.