An Energy Efficient Code Offloading Framework In Mobile Device Cloud

In recent years,mobile applications are becoming more and more complex,which requires mobile devices upgrade hardware to provide higher processing performance.Existing researches have proposed to offload application codes to nearby mobile devices to improve application performance where these nearby mobile devices are called mobile cloud.However,these mobile devices are also powered by battery with limited capacity,the existing code offloading schemes do not consider the impact of CPU frequency on computation energy consumption,an unreasonable code offloading scheme will cause significant energy consumption.The limited battery capacity is still the bottleneck of the performance of mobile devices,and it is of great theoretical and practical significance to propose a code offloading scheme with high energy-efficiency.Based on the characteristics of CPU frequency and energy consumption,this paper presents a high energy-efficiency application code offloading framework in mobile clouds.The theoretical analysis and experimental data verify the nonlinear relationship between mobile device CPU frequency and computation energy consumption,and the lower the CPU frequency,lower energy consumption.According to the relationship between CPU frequency and energy consumption,a code offloading framework with high energyefficiency is proposed.The basic idea of our code migration framework is to divide the tasks on high performance devices into multiple parallel subtasks,then offload these subtasks to nearby mobile device operating at low performance status,as a result,not only energy is saved by reducing the energy consumption of subtasks,but also performance is guaranteed through the parallel execution of program.To maximize energy efficiency,we formulated the offloading problem into a mixed-integer nonlinear optimization problem and proposed a heuristic algorithm to solve the problem.To test the performance of our scheme,we use Mandelbrot set program as the test program,and the original program was divided into multiple sub-tasks and offloaded to nearby mobile device operating at different CPU frequencies.Experimental results show that compared with these scheme that did not relationship between CPU frequency and energy consumption,our scheme can save about 20% and up to 50% energy while ensure the same performance.