Research Of Network Virtualization In Software Defined Data Centers

As the underlying infrastructure of cloud computing, data center has draw more and more attention. Virtualization is the key technology in realizing cloud computing, it is widely used in data centers and promotes the advent and development of software defined data center. In a software defined data center, computing virtualization and storage virtualizaion have been researched for many years while the network virtualization technologies arestill in the early stage, many problems need to be investigated, one of which is the energy problem of the network. Energy problem of data centers has long been concerned by both industrial and academe. The power consumption of network devices account for 10%-15% of the power consumption of the whole data center. The advent of Network Function Virtualization(NFV) brought new chances for energy saving in data centers. When traditional hardwares are replaced by commercial servers and software, this part of power consumption may be reduced. However, as far as we know, no prior work has focused on the energy efficiency of network function virtualization. Therefore, in this paper, we tried to investigate the power consumption of different NFV implementations. We build a real-world platform and measure the power consumption of the server using real-world traffic. We analyzed the power of the same data plane when processing different traffic, the power of different data planes when processing same traffic, the impact of I/O virtualizations have on power, the power differences between different I/O virtualization technologies, power difference of VNF processing different traffic and the power differences of different VNFs processing same traffic. Our results can be used to guide the deployment of NFV to reduce power consumption by choosing proper data plane to process different traffic, choosing or designing the I/O virtualization technology that can achieve high performance with low power consumption.