Data And Computation Offloading In Hybrid Heterogeneous Networks

Recently,with the development of communication technology and the improvement of the application level of urban informatization,the construction of Smart City emerges as the times require.As an advanced form of the informationbased city,Smart City integrates urban composition systems and services based on the new generation of information technology such as the Internet of Things(IoT),cloud computing,so as to improve the efficiency of resource utilization and the quality of life of citizens.From intelligent transportation systems to real-time air quality monitoring,hundreds of millions of IoT devices are constantly generating data to support the operation of the Smart City.Moreover,emerging network applications(e.g.,mobile high-definition video,virtual reality/augmented reality(VR/AR))have generated huge mobile network traffic while enriching people's lives.Under the constraints of limited bandwidth,storage,and computing resources,traditional single network architectures cannot achieve the transmission,storage,and processing of these surges of data in real time,which can greatly reduce the operational efficiency of Smart City.Considering that Smart City is a hybrid heterogeneous communication network system based on wireless local area networks(WLAN),public mobile communication networks(e.g.,LTE,5G),and the Internet of vehicles(IoV)etc,there exist redundancies in the communication,storage,and computing resources among different network architectures.Therefore,the key to the sustainable development of the Smart City is to design a joint resource management mechanism suitable for the hybrid heterogeneous networks.Specifically,for data/computationintensive services,when the target network resources are limited,the vacant resources in other complementary networks can be applied for offloading,thereby achieving rapid response to the services.Among them,how to deal with the heterogeneity of networks,resources,and how to ensure the consistency of data and computing services are all urgent problems to be solved in the design of the offloading mechanism.In view of this,this paper focuses on data and computation offloading technology in hybrid heterogeneous networks.For the hybrid heterogeneous network consisting of LTE and WLAN,we study the “MVNO oriented Data Offloading Mechanism”.We first introduce the data utility models of MNO and MVNOs,and the data offloading problem is modeled as a noncooperative traffic inventory game.Then,based on the Cournot and Stackelberg models respectively,we study the inventory game under two MVNO business models(i.e.,peer-to-peer and non-peer),and we further analyze and prove the existence and uniqueness of the two equilibriums.Traffic assignment strategies for the two models are developed based on the genetic algorithm.Finally,through numerical results,it is analyzed and proved that the proposed data offloading mechanism can maximize the network utilities while migrating the data load of the cellular network.For the hybrid heterogeneous network composed of LTE and IoV,we investigate “Edge Caching assisted Offloading Mechanism for Video Streaming”.We first study the vehicle mobility model based on the communication time between the vehicle and RSU.Then,to ensure the consistency of the video streaming service,we design a pre-caching scheme according to the mobility of vehicles.Furthermore,considering the limited storage space of RSU,a QoE oriented video playback rate adjustment scheme is proposed.Finally,through numerical results,compared with the caching scheme without rate adjustment,this RSU-based edge caching mechanism is capable of meeting an additional user requirement by10% to 40%.For the hybrid heterogeneous network composed of LTE and IoV,we study“Vehicular Cloud assisted Computation Offloading Scheme”.The system model and network architecture of vehicular cloud are first presented,and the transmission,queuing,and processing of computing tasks are modeled.Then,in order to characterize the instability of onboard computing resources,we study a vehicular mobility model based on the dwell time of vehicles.Next,considering the inter-dependency of computing tasks,we formulate an NP-hard task scheduling problem with the objective of minimizing the response time.Based on the global movement information and the real-time location information of the vehicle,a Time Expanded Graph(TEG)and Deep Reinforcement Learning(DRL)task scheduling mechanism are proposed respectively.Finally,through numerical results,it is proved that the VC assisted offloading scheme can guarantee low task response time while alleviating the workload of the edge servers.