Research On Task Offloading Algorithm In Multi-cloudlet Cooperation Pattern

Mobile cloud computing(MCC)is a newly-developing Internet applied pattern that combines cloud computing with mobile Internet technology,the research on MCC frame model,application scenario and task allocation algorithm in mobile cloud computing environment is very important.Cloudlet pattern as a way to mobile cloud computing,it has the advantages of fast response,high network bandwidth and low transmission delay.But the focus of most research remains on “mobile device-one cloudlet-remote cloud” scenario,little consideration of framework design,task offloading,and offloading algorithm in “mobile device-multi-cloudlet cooperation” pattern.Therefore,designing a mobile cloud computing framework for multi-cloudlet cooperation pattern,makes full use of the advantages of cloudlet,and improves the service quality of mobile applications,which will have important applied value.To solve the problem of task offloading in multi-cloudlet cooperation pattern,in this paper,a multi-cloudlet cooperation mobile cloud computing framework based on YARN is designed.At the same time,we make the weight relation graph to model the application and divide the application offloading granularity,mobile cloud computing task offloading algorithms suitable for multi-cloudlet environment are proposed,providing optimal and approximate optimal offloading strategy for different types of applications to shorten task execution time and reduce energy consumption of mobile devices and satisfy the various needs of user.The main work of this paper includes:1.A mobile cloud computing framework for multi-cloudlet to execute offloading tasks is designed.The basic idea of cloud computing framework YARN is used,making it suitable for multi-cloudlet to execute offloading tasks from mobile devices.Introducing a high performance proxy server for each cloudlet in the multi-cloudlet system to manage and maintain their own,the proxy server is consist of application layer,cloudlet node management layer and mobile device layer,we put the YARN main function into the cloudlet node management layer,improving the reliability and stability of multi-cloudlet system.2.By constructing the weight relation graph,one unit of granularity is a single application component,modeling the application.Taking the local execution time ofmobile devices,local execution energy consumption,offloading time,transmission time and consumption into account with the weight relation graph,building application partition model for execution time optimization and energy consumption optimization and normalized weight partition,providing reliable parameters for task offloading algorithm.3.We proposed the task offloading algorithm for multi-cloudlet cooperation environment.Taking the computing power of mobile devices,calculation power,transmission power,network connection status,cloudlet acceleration rate,application complexity and other factors into account,aim to optimize task execution time and energy consumption of mobile devices.We proposed optimized branch and bound algorithm for time and energy consumption(OB&BATE)and multi-cloudlet particle swarm optimization algorithm(MCPSOA),the algorithms offer the optimal and near-optimal offloading strategy based on user demand.Simulation experimental results show that application weight relation graph could provide reasonable,flexible application partition model with execution time optimization,energy consumption optimization and normalization weight model.The OB&BATE and MCPSOA algorithm for multi-cloudlet cooperation environment can provide optimal and approximately optimal offloading strategy within a reasonable time,significantly reduce the task execution time,reduce energy consumption of mobile devices.