Low Energy Consumption Calculation Unloading And Cost Optimiaztion In Heterogeneous Environment

With the rapid development of cloud computing and the Internet of Things,a series of computationally intensive and delay-sensitive applications such as speech recognition,natural language processing,wearable devices,and online games continue to appear.However,smart mobile devices(SMDs)Battery,memory,CPU and wireless media resources are limited,and these applications cannot be processed efficiently.In order to solve the shortcomings of limited storage capacity and insufficient computing power of SMDs,mobile edge computing(MEC)can solve this problem.Partial computing offloading technology is used to intelligently schedule some tasks of applications from SMDs with limited resources to high-performance edge servers or cloud data centers(CDC)for execution.However,due to the limited bandwidth,it brings additional communication overhead and inevitably increases the delay of tasks in the process of offloading to the edge server.Therefore,how to strike a balance between high resource consumption and high communication costs,provide users with low-latency,energy-saving services,and further reduce the cost of the system,is a very challenging topic.The purpose of this paper is to optimize the energy consumption and cost of the system,and research on the computational offloading strategy and resource allocation of SMDs.The main research contents and innovations are as follows:First,this paper constructs a partial computing offloading model based on the edge(edge-end binary architecture model).It is assumed that the tasks of SMDs can be executed in parallel on SMDs and edge servers.It comprehensively considers the tasks performed on SMDs and edge servers.Execution time,and data transmission time.It also considers the latency limit,CPU speed,transmission power limit,the available energy of SMDs,and the maximum number of CPU cycles and memory in edge servers.By jointly optimizing task offload rate,SMDs CPU speed,available channel allocation bandwidth and transmission power of each SMDs in each time interval,the total energy consumption of SMDs and edge servers is minimized.Secondly,this paper adds a cloud data center based on the above binary model to construct a partial computing offload model under cloud-edge collaboration(cloudedge-end three-tier architecture model),which mainly uses queuing theory to perform the offloading process in the model.In-depth research on issues such as energy consumption,execution delay,and payment cost,and also consider the limiting factors such as delay,CPU speed,transmission power,etc.In this model,when the number of tasks is large,the task processing capability on the edge cannot cope with the tasks that arrive.However,tasks that exceed the processing capacity of the edge will be offloaded to the CDC for execution.At the same time,a new measurement method is introduced to consider the characteristics of different devices.Edge servers and CDC pay more attention to the utilization of their processors,while SMDs are more Pay attention to its energy consumption.Therefore,the cost of SMDs in this metric is expressed in terms of energy consumption and evaluates the cost of edge servers and CDC by the number of tasks performed.Finally,the total cost in the system is minimized by jointly optimizing the task offload rate,the allocated bandwidth of the available channels and the transmission power of each SMDs in each time interval.Finally,according to the above two models,a nonlinear constrained optimization problem is proposed,and a new hybrid meta-heuristic optimization algorithm,Genetic Simulated-annealing-based Particle Swarm Optimization(GSP)is adopted to solve the two problems proposed,and get the approximate optimal solution.The GSP algorithm can realize the joint optimization of each part of the binary architecture or the three-tier architecture in the computing offloading process,as well as the dynamic resource allocation of the base station or CDC.Experimental results show that the algorithm can achieve lower energy consumption under the binary architecture model and lower system cost under the three-tier architecture model in a shorter convergence time.