Traffic Aware Virtual Machine Packing with Link Load Constraint in Cloud Data Center

• Main Authors: TSAI, JUNG-TSE, 蔡鎔澤
• Other Authors: HOU, TING-CHAO
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/t55g77

碩士 === 國立中正大學 === 通訊工程研究所 === 106 === In recent year’s cloud network, virtualization in data centers is an efficient way to increase resource utilization and reduce energy consumption. Data center mainly uses virtualization to provide different services to cloud tenants. When the cloud tenants rent for a virtual machine, it would ask the data center for computing capacity and the inter-VM bandwidth. Therefore, the cloud provider would provide the basic capacity and bandwidth according to tenant’s demand. As the technology advances, cloud tenants are renting more and more virtual machines. This situation will cause the data center to use excessive servers to provide external services and bring unnecessary electricity consumption to the data center. Therefore, some methods like virtual machine packing are proposed to reduce the usage of data center servers, thereby reducing unnecessary electricity consumption and making efficient use of servers. The methods proposed previously only aimed at the situation where the server's incoming and outgoing data traffic is limited by the transmission rate of the network interface. However, the connection between servers or between virtual machines in the data center should also consider the path capacity. Once the capacity of the path is taken into consideration, the link load on the path becomes particularly important. In this thesis, we use Equal Cost Multipath (ECMP) to do the path routing. Comparing with the shortest path routing, ECMP can effectively relieve the overload on the specific link, thereby improving the efficiency of virtual machine packing.