Researches On Service Resource Allocation In Digital Twin-driven Edge Computing Network

The edge computing networks extends computing services from the cloud center to the network edge closer to users.By rapid collection of network parameters and efficient allocation of various service resources,edge computing networks provide users with a variety of application services with high reliability and low latency.However,with the diversification of service and the expansion of scale,it is harder to timely acquire parameters and conduct efficient management of service resources,in which traditional resource allocation methods show great limitation.Meanwhile,digital twin(DT)provides a new technological approach for lifelong monitoring,flexible optimization and real-time control of physical objects in a distributed way,which is highly compatible with edge computing networks,thus helping solve the problem above.Therefore,this thesis introduces DT technology to conduct in-depth research on the optimization problem of multi-service caching and heterogeneous resource allocation in edge computing network,with the following contributions:(1)Facing the requirement of low latency updating for DT model of devices,this thesis proposes a multi-edge collaboration-based low-latency DT model updating framework,and constructs a joint optimization model of edge association,bandwidth and computation resource under quality of service(Qo S)constraints.Then,the model is decoupled into two sub problems: edge association optimization and resource allocation optimization.We propose an alternative optimization-based cluster formation algorithm to solve the problems.Simulation results show that our proposed scheme reduces the latency of DT model updating by around 62% and 35%,compared to the distance-based grouping scheme and the stochastic clustering scheme respectively.(2)To meet the requirements of efficient caching of DT models towards multi-service scenario,we use network slicing to divide the DT models of network devices into several accessible contents(ACs)corresponding to different service types.For each AC,we design a multi-service-oriented caching scheme,which is subsequently transformed into a Stackelberg-coalition game model.Then,we propose a coalition formation-based distributed caching algorithm.Simulation results prove that our proposed scheme,compared to clustering-based caching scheme and non-cooperative caching scheme,improves the total system throughput by at least 12% and 2.5 times respectively,and the total utility of IESs increases by over 14% and 2.3 times respectively.(3)To cope with the problem that heterogeneity of edge computing network resource aggravates supply-demand mismatching,we propose a DT-based heterogeneous resource allocation scheme,which promotes resource cooperation among IESs by formulating a coalition game model.In each coalition,intelligent edge servers(IESs)maximize their utilities by sharing and jointly optimizing multiple resources to match resource service requests of terminal devices(TDs).Simulation results show that the proposed scheme exceeds grand coalition scheme on improving IES average utility by around 65%,which is more adaptable to large-scale applications.Based on the characteristics of DT driven edge computing network,we design a series of distributed optimization and decision schemes,which provide ponderable reference for researches on DT-based service resource allocation in complex edge computing scenarios.