Research On Computation Offloading Strategy Based On Mobility Behavior Analysis In Distributed Mobile Cloud Computing

With the rapid growth of intelligent terminals and the rapid development of the mobile application market, the performance (storage,computing power) limitation of intelligent terminals and the real-time requirements of some applications make traditional cloud services increasingly unable to meet the needs of mobile users. In this case,distributed MCC (mobile cloud computing) becomes a promising solution to solve these problems. Cloudlet, as one of the service provision mode of distributed MCC, pushes the computing resources to the edge of the network. But cloudlet resources are limited and its coverage is small, and the stability of the connection with the terminal is influencedby the user's movement and its capacity. Therefore, how to utilize the mobility characteristics of the users to optimize the distributed computation offloading, i.e., to improve the success rate of offloading and to reduce the terminal energy consumption, is very important when multiple cloudlets can be accessed.As the important basis of the work in this paper, the related technologies are introduced, including cloud computing, mobile computing and mobile cloud computing, as well as the related technologies of human mobility behavior research, such as the law of human mobility, The human mobility model and the mobile location prediction method. A tail-matching algorithm based on user's wireless access sequence is designed for wireless access point prediction.Experiments based on reality mobility dataset are conducted to prove that the algorithm can effectively predict the wireless access point. And then the cloudlet reliability evaluation is defined base on the human mobility feature, the cloudlet load and the access time. GADCO (Genetic Algorithm for Distributed Computation Offloading), an improved computation offloading strategy based on Genetic algorithm is proposed. GADCO is expected to reduce terminal energy consumption and improve the offloading success rate of work flow under the premise of satisfying user's latest completion time. Finally, experiments are conducted for performance evaluation, comparing GADCO with different existing computation offloading algorithms. The experiment results show the performance advantages of GADCO in computation offloading.