Component-based Mobile Cloud Computing Architecture And Performance Optimization

In recent years,the number of smart mobile devices has grown significantly and their performance has increased in terms of computing capacity,storage capacity,bandwidth,and sensors.More and more applications are developed and provided to mobile devices.Since the performance of mobile devices cannot be greatly improved due to its resource constraints,a new computing paradigm called mobile cloud computing has come into the people's vision.Generally,mobile cloud computing refers to the information and service access pattern of obtaining scalable infrastructure,hardware devices,software platform,application and service on demand through wireless access network by mobile users.In the state-of-the-art mobile cloud computing architectures,Cloudlet-based computation offloading architecture has the advantage of making full use of network edge computing resources,reducing the latency of user access to computing services and reducing core network bandwidth consumption.In this thesis,Cloudlet-based mobile cloud computing architecture and its performance are investigated,and the main research work is presented as follows:Firstly,a component-based Cloudlet computation offloading cluster dynamic networking architecture called PMC2O is presented,including Cloudlet dynamic discovery,the comparison of computing capability,joining and remove of devices as well as Cloudlet merge process.In order to implement the dynamic self-management of component-based Cloudlet computation offloading cluster,the Proactive Remove Decision Algorithm based on Computation Resource Sharing Degree(PRDA-CRSD)is presented.Based on the mobility of devices in the Cloudlet,an improved Simulated Annealing algorithm named SA-UM is proposed to optimize the offloading strategy of component.The self-management performance and the dynamic networking behavior of Cloudlet are evaluated based on the above-mentioned dynamic networking mechanism by OPNET.The simulation results verify the effectiveness of the proposed mechanism and algorithms.Secondly,considering the mobility of nodes,a new metric to illustrate the delay of mobile computation offloading called re-running delay is proposed,which refers to the additional offloading delay caused by the re-execution of the offloading failure component.The effect of method of loads,computation capacity of devices and user mobility to the re-running delay is evaluated by system level simulation.