Computation And Data Offloading In Mobile Edge Computing Networks

With the continuous development of mobile Internet,Internet of Things,artificial intelligence and other technologies,new mobile applications and network services continue to emerge,such as virtual reality,augmented reality,ultra-highdefinition video,autonomous driving,interactive games.While these emerging applications and services have enriched people’s lives,they have also caused a dramatic increase in computation demand and data traffic in mobile communication networks.The emergence of Fifth-Generation(5G)mobile communication networks and mobile edge computing(MEC)has promoted the rise of mobile edge computing networks,which has become an important way to improve the computation processing capacity and content distribution efficiency of mobile communication networks.The mobile edge computing network includes central cloud,edge clouds and terminal devices.It is not only the continuum of heterogeneous “device-edge-cloud” network,but also the continuum of heterogeneous resources such as communication,computation,and storage.Currently,mobile edge computing networks are facing an unbalanced distribution of service requests and resource supply,causing partial overloads of computing and data services.Offloading technology has become the key technology to improve network service quality and user service experience.Due to the characteristics of network heterogeneity and resource heterogeneity of mobile edge computing networks,offloading technology faces many problems and challenges.Firstly,although 5G and mobile edge computing have respectively improved the communication and computing capabilities of the network,the current cloud radio access network(C-RAN)and edge servers are independent of each other,and heterogeneous resources such as communication and computing are configured separately,resulting in low utilization of network resources and poor system energy efficiency.The dense deployment of 5G network also makes users suffer more complex communication interference during offloading.Secondly,because the mobile edge computing network has a heterogeneous “device-edge-cloud” network architecture,current offloading strategies can only achieve simple cooperation and are insufficient to complete complex tasks.At the same time,due to the coupling of computation tasks and network dynamics,the energy consumption,delay,and resource cost of offloading complex computation tasks are increased.Finally,although device-to-device(D2D)communication and caching are important ways to assist mobile edge computing networks in data offloading,due to the timeliness of contents and the self-interest of users,the data offloading cost of mobile network operators continues to rise,which brings obstacles to the popularization and application of mobile edge computing networks.To deal with above problems and challenges,this paper studies computation and data offloading technology in mobile edge computing networks.The main research contents are as follows.Aiming at the problem of weak integration of mobile edge computing and wireless access network,and the poor system energy efficiency,this paper studies “Computation Offloading and Multidimensional Resource Management Technology for Integrated MEC and C-RAN”,and proposes a MEC and C-RAN integration scheme to realize device-edge collaborative computation offloading,and computation and wireless resource management.Combined with network function virtualization technology,the unit energy consumption of the system is reduced.Using Lyapunov optimization theory and divides the unit energy consumption optimization problem is divided into four sub-problems.Then,the primal and dual decomposition technology is used to obtain computation offloading and elastic computation resource scheduling policies.Matching game and geometric programming are designed to solve subchannel and power resource allocation.The algorithm proposed in this paper can support online distributed implementation,so as to achieve the goal of reducing energy consumption with low time complexity.Simulation results show that the proposed integration of MEC and C-RAN,as well as computation offloading and resource management method,can reduce system unit energy consumption and task completion delay by 59% and 57%,respectively.To solve the problem of poor device-edge-cloud collaboration and high computation cost of complex tasks,this paper studies “Computation Offloading and Resource Scheduling Technology for Cloud-Edge Hybrid Networks”,and proposes two device-edge-cloud collaborative computation offloading and resource scheduling algorithms under different system conditions.These two algorithms fully consider the mobile applications with task call graphs,time-varying wireless channels and available computation resources,and different prices of edge clouds and central cloud,and effectively reduce energy consumption,latency and computation resource rental costs for computation tasks.Specifically,with full system information,a computation offloading and resource scheduling scheme based on semidefinite relaxation and dual decomposition is proposed.With partial system information,a deep reinforcement learning framework is designed,and an improved deep deterministic strategy gradient algorithm is proposed to learn the computation offloading and resource scheduling policy with a mixed action space.Simulation results show that the two methods have good convergence performance and can reduce the energy consumption,delay and resource cost of computing tasks by at least 15.5%.In order to solve the problem of high cost and poor information freshness during content distribution in mobile edge computing networks,this paper studies “Device-to-Device Content Caching and Updating Technology in Mobile Edge Computing Networks”.The data traffic in mobile edge computing networks is offloaded by caching and device-to-device communication.Combined with the information freshness constraint of cached content,the utility maximization problem of mobile network operators is formulated.This paper presents a twin-greedy based content caching algorithm and a Gibbs-sampling based content updating algorithm to ensure the timeliness of content and improve the data offloading utility of mobile network operators.Considering mobile users’ self-interest,this paper also designs a payment determination mechanism based on reverse auction,which satisfies the properties of individual rationality and truthfulness.Simulation results show that the proposed mechanism can improve the data offloading utility for mobile network operators by at least 17.5% while guaranteeing the information freshness of cached contents.