Research On Task Offloading Technologies In Mobile Cloud Computing

Along with the continuous development of mobile communication technologies,the mobile internet business has been booming.Meanwhile mobile terminals represented by smart phones,tablet PCs,mobile terminals have been widely popular,and provide a strong support for the rapid development of mobile applications.Mobile applications have entered into various areas of people’s lives,and people can get a variety of information and services through the mobile internet currently.Although mobile terminals have made great progress in hardware technology,their resources are still limited compared to traditional computers.With the rise of cloud computing technology,more and more enterprises and individual users begin to use cloud services to improve work efficiency and reduce costs.Cloud service providers can provide users with resources through the network in an on-demand,extensible way.Mobile cloud platforms mainly consist of central cloud,cloudlet and ad hoc virtual cloud.Mobile cloud computing combines the respective advantages of cloud computing and mobile Internet,and can better meet the needs of mobile terminal users in terms of computational efficiency and convenience of services.Task offloading in mobile cloud computing is to move the tasks with heavy load from mobile terminals to the cloud platform,thus breaking through performance bottleneck brought by resource limitations of mobile terminals.In this paper we mainly study the problems of task offloading strategy and performance optimization for task offloading in mobile cloud computing.Specific work is as follows:First,the drawback of existing task offloading strategies is analyzed(i.e.,resource characteristics of a single cloud platform can not meet the offloading requirements of various tasks),and we propose the interconnection architecture of mobile cloud platforms.This architecture aggregates various resources on heterogeneous mobile cloud platforms(central cloud,cloudlet and vehicular cloud),thus forming resource supply mode of multi granularity so that various requests for task offloading can be supported.To improve task offloading success ratio and offloading performances,we propose a flexible offloading strategy which is able to select a suitable cloud platform based on offloading performance analysis.For vehicular cloud,an algorithm is proposed to select the reliable worker node according to the connection status and resource status of the worker nodes.Experimental results show that when the central cloud and the cloudlet can not meet the offloading requirements,vehicle cloud can save about 46% of energy consumption of the intelligent terminal,and the average task offloading success ratio is about 85%.Then,for tasks with heavy load on vehicular terminals,the limitation of central cloud is analyzed(i.e.,larger network delay).To reduce task response time,we propose a distributed cooperative offloading mechanism that combines cloudlets and a central cloud.According to whether the requested tasks can be executed in parallel,the tasks are divided into two categories: un-parallelizable tasks and parallelizable tasks.For un-parallelizable tasks,we propose various efficient aggregation schemes to meet different task requirements.For parallelizable tasks,we propose efficient task allocation and cooperative execution mechanism,i.e.,a certain amount of data in cloudlets is processed by the central cloud,so that the distributed and parallel computation can be implemented to reduce task response time.Experimental results show that the proposed mechanism can reduce about 45% of average task response time of the existing method.Next,due to the strong mobility of vehicular terminals,vehicular cloud mainly uses position information to select the nodes(vehicular terminals)to provide services.We propose a contention-based node adaptive position update approach to reduce the response time of vehicular cloud.The proposed approach includes three algorithms,i.e.,adaptive position update for next hop algorithm,deleting unreachable next hop algorithm and broadcasting contention-based beacon algorithm.The approach adaptively triggers position update for next hop based on position deviation and data transmission.Moreover,the approach updates the position information of the key nodes that affect network connectivity and data transmission.The experimental results show that compared with the existing methods the proposed method reduces the average response time and improves the task delivery ratio.Last,due to the limited wireless resource of AP connected to cloudlet,when multiple users are using offloading services in the cloudlet simultaneously,how to ensure that the response time of multiple users meets their own requirements is difficult.We propose a transmission scheduling algorithm based on offloading service utility in the cloudlet.The proposed algorithm considers the fairness from the perspective of services provided by the cloudlet.The proportional relation among data rate,service ratio and task remaining time is employed to establish a service utlity function.The proposed algorithm selects the user corresponding to the maximum utility value at each time slot in order to improve service fairness,quality of service and service efficiency.Experimental results show that the proposed algorithm can increase about 3.7% of average system service success rate and about 10% of average system service efficiency corresponding to the existing method.