Network-Constrained Virtual Machine Deployment Strategy in Open Cloud Platform

• Main Authors: Feng-yi Chou, 周鳳儀
• Other Authors: Jiann-liang Chen
• Format: Others
• Language: en_US
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/04044364346112427036

碩士 === 國立臺灣科技大學 === 電機工程系 === 100 === Cloud computing emerges as a novel application service layers. It provides diversity of applications and virtualization resources, such as storage services, CPU, and memory. Thus, when it is compared with the traditional data center, cloud data center becomes very attractive solution for enterprises where enormous management cost could be reduced. Virtual machines hold important role for cloud data center service. Since the data center may be distributed over many regions, and the network conditions within cloud data center is different, so the deployment of virtual machines will influence service performance obtained by users. This study implements Eucalyptus cloud computing platform as a cloud data center system, and proposes network utilization constrained scheduling algorithm that works according to the current network status between users and associated node controllers for deploying the services. In addition by OpenFlow switch is implemented to manage the intended flow within users and virtual machines. The experiment is conducted upon node controller in the same and different subnet. For each experiment two scenarios are performed, First scenario is conducted within normal network condition, second scenario is considering bottleneck event at node controller. The experiment results indicate that in the same subnet, the virtual machines deployed with network utilization constrained scheduling algorithm can improve the transmission time 2.63% and 68.72% compare to traditional Round Robin and Greedy algorithms consecutively. In considering bottleneck event, the proposed method can improve 32.79% and 60.24%. In the different subnet scenario, the proposed mechanism improves 21.99% and 52.10%, eventually with bottleneck event the system is improved as 22.74% and 49.86%.