Research On Optimal Resource Allocation And Deployment Technologies In Edge Computing Environments

As the number and types of user terminal devices increasing,more and more complex Io T applications cannot be processed in a real-time and cost-effective way,which has become a bottleneck restricting the development of their business.On the one hand,due to the emergence of new application fields of the Internet of Things(Io T),such as virtual reality,augmented reality(VR/AR),ultra-high definition video broadcast and intelligent manufacturing,the new complex,diverse and real-time business requirements have been put forward.On the other hand,the current Io T terminal devices have shortcomings such as insufficient processing capacity and limited battery capacity,making it difficult to provide real-time processing for complex and diverse Io T applications.The emergence of edge computing services can improve local data processing capacity,reduce data transmission delay and equipment cost to a certain extent,and provide effective solutions for these applications.How to provide the optimal or better resource allocation strategy and deployment scheme for the complex and diverse Io T services near to the edge of the network for Io T or data sources is a key scientific problem to be solved urgently in edge computing environments.These teconologies,sush as computational offloading,resource provision,cache content placement and edge server deployment are the basis of localized data processing and optimized resource allocation,and their execution efficiencies and execution costs will directly affect the overall performance of the edge computing system.Therefore,considering the characteristics of Io T application delay sensitivity,high computational intensity and heterogeneity of of three-layer device-edge-cloud resources,this dissertation focuses on computation offloading,resource provisioning,cache content placement,edge server deployment,and systematically studies optimal resources allocation and deployment of edge computing for varied complex services,that promote system performance,the quality of services and user experience,the main research work and contributions of the dissertation are as follows.(1)It is studied that a computation offloading strategy for multi-component applications in edge computing.At present,the processing power and battery power of the user terminal are limited.The computation offloading technology can be used to transfer the Io T real-time computing tasks to edge servers or a remote cloud.How to expand the processing capacity of user terminals and meet the needs of large-scale real-time Io T applications is an urgent problem to be solved.This dissertation proposes a computation offloading strategy for multi-component applications in edge computing.This strategy takes a component as the application division unit,considers the behavioral characteristics of application components,uses query graph and data label graph to describe.and constructs the fuzzy similarity matrix.It determines the clustering relationship according to the membership between components,uses the fuzzy clustering algorithm to cluster components,and realizes the accurate division of high cohesion and low coupling for Io T applications.Then the dissertation comprehensively considers the access delay and energy consumption,calculates the overall cost of task distribution to local and edge or cloud nodes,and analyzes the user location,the context of computing,storage,and network resources in edge computing.Once the offloading conditions are met,the dynamic subgraph matching algorithm is adopted to perform computing offloading.The experimental results show that the proposed algorithm reduces energy consumption of user equipment and the service delay,improves the execution efficiency and user experience.(2)It is presented that the energy-aware edge-cloud multi-level dynamic resource allocation method.Under the constraints of energy consumption of terminal equipment and performance of edge server,how to reasonably utilize the distributed multi-layer heterogeneous resources composed of cloud,edge servers and terminal equipment,and minimize energy cost while meeting the deadlines of applications is a key issue.This dissertation proposes an energy-aware multi-layer dynamic smooth resource allocation strategy in a collaborative edge and cloud system.Firstly,the Weighted Voronoi Diagram(WVD)is used to determine the service area of the edge server.The AR(p)model is used to predict each edge server workload and select the locations of sites according to the history workloads.Then,according to the energy consumption cost of the whole system,the reconfigurable resource smooth allocation problem is transformed into a multi-dimensional knapsack problem(MKP).Using the energy-aware Greedy algorithm and the dynamic node management strategy,the optimal resource fair allocation scheme with the minimum energy consumption cost while meeting the deadline is finally obtained.The experimental results show that the proposed method outperforms the comparative algorithms in terms of energy consumption and SLA violations.(3)It is designed that a distributed collaborative edge cache placement algorithm for edge computing.A single edge server has limited storage capacity in edge computing systems,and the hierarchical caching mechanism causes a long delay and wastes storage space and a cache content placement policy considers only a single factor causes a decrease in cache value and higher bandwidth consumption.This dissertation designs a distributed collaborative cache placement algorithm in edge computing.With the distributed collaborative cache architecture,the user set within the coverage of each edge server is first determined,and then the data access cost is calculated by the distance between the cache server and the terminal device,the content popularity,and the cache content size.Then minimization problem on the access delay cost is reduced to 0-1 integer linear programming problem.The meta-heuristic IT?algorithm is used to search for an optimal cooperative cache placement solution.Compared with other traditional content placement algorithms,the experimental results show that the proposed data placement strategy ECCDP＿IT? can efficiently search the best placement of popular data,improve cache hit rate,and reduce back-end network transmission traffic as well as increase user satisfaction.(4)It is recommended that cost-aware deployment optimization strategy for edge servers.How to select the location of edge servers and determine the number of servers at this location to achieve the tradeoff between the low latency and high node utilization is an urgent problem to be solved.This dissertation proposes a cost-aware deployment optimization strategy for edge servers.Firstly,the resource allocation ratio is calculated by using the user association matrix and the resource allocation matrix,and then the total delay of the regional request is calculated by the resource allocation ratio,the regional average load,and the access delay between the edge sites.Finally,the minimum objective function based on site cost and total access delay cost is established,and the server deployment problem is reduced to Mixed Integer Nonlinear Programming(MINP)problem.The Benders decomposition algorithm is used to find the optimal locations and quantity of edge servers.The simulation results show that the edge server sparse deployment optimization strategy based on Benders decomposition achieves a lower overall deployment cost of the edge server while achieving the tradeoff between low access latency and higher resource utilization.