Aiming at the solidity of code and the difficulty of upgrades of existing Virtual Data Center (VDC) management platform, a VDC management platform based on Software Defined Network (SDN) was proposed. The proposed platform was composed of VDC Management subsystem (VDCM), VDC Computing Resources Control subsystem (VDCCRC) and VDC Network Resources Control subsystem (VDCNRC). A loosely coupled architecture was built by RESTful API interaction between subsystems. VDCNRC managed data center network by SDN controller, VDCCRC managed computing resources of data center by open source cloud computing platform, and a VDC management algorithm framework was built in VDC management subsystem to develop rapidly VDC management algorithm suitable for production environment. By utilizing Mininet, Openstack and Floodlight to set up test environment, the results show the proposed platform can support running, migrating or deleting virtual machines by Openstack, implement bandwidth resource isolation between VDCs by Openflow controller, and support the operations of creating, deleting, or updating VDC.

By introducing the current research progress of Virtual Data Center (VDC) embedding, and in accordance with the reliability requirement of VDC, a new heuristic algorithm to address reliability-aware VDC embedding problem was proposed. It restricted the number of Virtual Machines (VMs) which can be embedded onto the same physical server to guarantee the VDC reliability, and then regarded reduction of the bandwidth consumption and energy consumption as main objective to embed the VDC. Firstly, it reduced bandwidth consumption of data center by consolidating the virtual machines, which had high communication services, into the same group and placed them onto the same physical server. Secondly, the consolidated groups were mapped onto the powered physical servers to decrease the number of powered servers, thus reducing the power consumption of servers. The results of experiment conducted on fat tree topology show that, compared with 2EM algorithm, the proposed algorithm can satisfy VDC reliability requirement, and effectively reduce a maximum of 30% bandwidth consumption of data center without increasing extra energy consumption.