Research On Dynimic Offloading Decision In Mobile Cloud Computing Systems

With the widely used of mobile devices,the amount of mobile data is explosively increasing.While the development speed of mobile device hardware is far less than the development speed of software,such as poor battery life,limited computing resources and storage resources.These issues will affect the user experience and limit the development of mobile applications.The mobile cloud computing can effectively solve the above problems by migrating the tasks of the device to the cloud as needed so as to reduce the energy consumption of task processing and shorten the delay of the computation intensive application.Offloading is the core technology in mobile cloud computing.However,offloading is not always beneficial to users.Device status and network condition will affect offloading performance.So how to make a effective offloading decision becomes a key issue.This thesis focuses on different scenarios,considers the different decision-making factors,and conducts a detailed study on how to make effective dynamic offloading decisions in mobile cloud computing networks.First of all,this thesis investigates the research background and the significance of the topic.Whereafter it expounds the challenges of mobile cloud computing.According to the existing research on mobile cloud computing,three appling architectures of cloud computing systems are summarized.Each architecture and its advantages and disadvantages are explained and analyzed.Combining the task offloading process,the structure introduces the function of each module and the flow between multiple modules from one to another by using the client-cloud structure.Four aspects of the existing mobile cloud computing offloading decision-making research are summarized.They are introduced with relevant research content,and analyzed with the innovation points and deficiencies of the research by the classified according to the optimization obj ectives.Secondly,for the problem of delay sensitive timely application of large amount of backhaul data,a dynamic offloading decision model for multi-users in heterogeneous network is proposed.In the system model,the heterogeneous characteristics of the wireless network and the cloud resources are considered.The user can offload the task through the Wi-Fi or the cellular network.The central cloud and the cloudlet can be used as the offloading location.Due to the limited computing resources of the cloudlet,the resource allocation problem is also considered.In addition to the user experience,the cost of user is also considered as one of the factors affecting offloading.Communication resources and computing resources are considered.Since the problem raised in this thesis is NP-hard problem,genetic algorithm can effectively solve such problems.Therefore,an improved genetic algorithm is proposed to find better offloading decisions.The simulation results show that the scheme can obtain a locally optimized offloading scheme.Finally,the offloading decision will be affected by many factors,such as user mobility.The network will be unstable or even the connection will fail resulting in the interruption of the offloading process and affecting the user experience.Therefore,under the distributed cloudlets architecture,this paper establishes a dynamic offloading model based on multi-factor perception.The computation-intensive application is divided into multiple tasks.The task is offloaded according to the connection probability,the availability of cloudlets,the channel condition,and the offloading energy coinsumption.Reduce the probability of connection failure while ensuring processing efficiency.Since multiple tasks are related to each other and each of them selects an appropriate offloading position to maximize the overall utility,the problem is solved using a coalitional game.The simulation results show that the proposed dynamic offloading decision scheme can obtain a better offloading scheme.