Research On The Task Offloading Strategy Based On Mobile Edge Computing

With the rapid growth of the Mobile Internet and the Internet of Things, the future network is faced with the challenge of higher experience rate, greater bandwidth, lower latency and higher reliability. To solve these challenges, mobile edge computing (MEC), which provides data computing, content delivery, and service awareness to mobile users,becomes a new network architecture in future 5G networks that will enhance the ultimate user experience. MEC provides the IT service and cloud computing capability to the wireless access network and sinks the service platform to the edge of the network. MEC greatly reduces the delay of service delivery and its cloud servers also provide computing and data storage capabilities to mobile users, which is ideal for task offloading.The task offloading technology can solve the problem that mobile terminals have low computing power and the limited battery capacity by offloading complex computing tasks from the mobile device to the cloud server through the wireless network. Because of MEC server in close proximity to mobile subscribers, it can provide users with great computing power and convenient access. This makes the task offloading path shorter and reduces offloading latency. Under the mobile edge computing system,the traditional task offloading strategy can not take advantage of the potential of MEC in the face of new features such as faster transmission rate and limited computational resources of MEC servers. Therefore, the thesis mainly studies the task offloading strategy under the mobile edge computing environment. The main work is as follows:First, the thesis presents a fine-granularity offloading policy in single user case. In MEC scenario, since mobile users are provided with a short-distance and high-rate access to cloud servers, the delay and energy consumption of uploading and downloading between mobile devices and cloud servers can be significantly reduced. This enhances the effect of fine-granularity contextual parameters inside the mobile application, such as computational intensity and exchanged data of each task, on making offloading decisions. So a fine-granularity mobile application partition model is established. Then the thesis constructs an optimization problem of minimizing the energy consumption while satisfying a strict delay constraint. To solve the problem with a low computing load, the thesis adopts binary particle swarm optimization (BPSO) algorithm. Simulation results show the proposed offloading policy significantly saves energy consumption while delay constraint is satisfied.Second, the thesis presents a distributed task offloading strategy in the multi-user case. Since the finite computation resources of MEC server and severe inter-cell interference limit the scalability of offloading, the thesis introduces the interference of wireless channel and delay of waiting for the virtual machine in the task offloading model. Then the thesis proposes a sequential offloading game approach to resolve the multi-user offloading problem and present a distributed task offloading decision-making program. Numerical results show that our proposed algorithm can achieve efficient performance in terms of both the experienced latency and energy consumption and scale well as the network system size increases.