Abstract
With the evolution of network technology, it is not only everyone is equipped with a smartphone, but also more and more smart devices are developed. There is going to step into the “Internet of Things” in the next generation. Under massive equipment connected to the Internet, the conventional core and access network infrastructure will not be able to continue to fulfill the requirement of network service. And to reduce the network deployment (CAPEX) and operating costs (OPEX), operators turn their path toward the integration of cloud platform and radio access network infrastructure. In this thesis, we take an open source LTE software project, OpenAirInterface (OAI) [1] as software radio access network application. And deploy OAI with virtual machine (VM) and Docker Container managed by OpenStack [2] which provides a real-world testbed for future fifth generation in assistance to network function virtualization management and deployment. We will use profiling tool to find the functional modules of eNodeB with heaviest system usage and then split them out to the other instance so that we can provide the least system resource to support maximum number of eNodeBs. By the experiment result we can learn that an eNodeB can offload about 12% ~ 25.5% CPU usage. Consequently, we utilize three resource allocation algorithms, Next-Fit, First-Fit and First-Fit Decreasing and show that the First-Fit Decreasing algorithm can bring the highest benefit which can improve about 20% resource utilization.