Research On Computing Power Coordination Technology In Edge Computing Network Based On Atomized Functional Components

The essence of edge computing technology is to deploy edge servers at the edge of the network to provide computing resources and network resources to users nearby.It has received extensive attention and has been incorporated into the 5G standard as a core technology.The concept of a computing power network has driven the change of the traditional Internet architecture and promoted the gradual transformation of the edge computing network towards the integration of computing into the network.Moreover,the development of microservices and containers has decoupled the traditional client-server model.Server-side applications are decoupled into atomized functional components that are uniformly scheduled by the gateway,and the computing power in the network is also encapsulated in these atomized functions in the component.However,from the perspective of completing various computing tasks for users,the atomized functional components with computing power embedded in the network are not peer-to-peer but are affected by distance,network conditions,computing resources,and functional components types,etc.The application requirements of various users are different,which leads to the ever-changing business requirements,and a business often needs multiple functional components to work together.Therefore,in the edge computing network based on atomized functional components,this thesis customizes the best coordination strategy for the tasks according to tasks characteristics and network status with considers the user’s demand for resources and the edge computing network’s provision of resources to provide users with services with the shortest latency.The specific research contents and contributions of the thesis are as follows:(1)An abstract model of distributed heterogeneous resources in edge computing network based on element graph is proposed to realize the centralized management of computing resources and network resources in edge computing network.(2)Different tasks have different computing resource requirements,network resource requirements and subtask dependencies,which lead to huge differences in computing coordination strategies customized for different tasks.How to quickly measure the latency performance of various computing coordination strategies is the top question to be addressed when developing a strategy.As for this,the thesis designs a service latency evaluation model,which can be used to quickly evaluate the service delay of various computing power coordination strategies.(3)Since formulating computing power coordination strategies is an NP-hard problem,the thesis studies the use of greedy algorithms,heuristic algorithms,machine learning,and subgraph isomorphism to formulate computing power coordination strategies.Furthermore,we conduct simulation experiments to compare their performance.And,for reducing the complexity of subgraph isomorphism,an adaptive adjustment strategy is proposed to reduce the search space and speed up the solution of the subgraph isomorphism algorithm.