Research On Resource Collaboration And Adaption Mechanisms In Smart Identifier Networking For Edge Computing

With the proliferation of network devices and traffic,edge computing has become a key technology to improve service quality and enhance users' experience in nowadays network environments.However,due to the "static and rigid" original shortcomings,the traditional network technology is incapable of supporting the efficient operation of edge computing because of its inefficiency in cross-network resources collaboration and customized service guarantees.For this reason,the design of edge computing networks based on the newly emerged network architectures and technologies has become a consensus in related research fields.As a new type of network architecture,the Smart Identifier Networking(SINET),which has significant advantages in the aspects like resource adaption and network control and management,can provide an ideal architecture for the efficient operation of edge computing.To this end,this dissertation relies on the design of SINET and copes with the resource collaboration and adaption issues of edge computing,focusing on solving the problems of task offloading performance evaluation,transmission and storage resource collaboration,and customized service resource provision.The main work and innovations of this dissertation are as follows:(1)To address the problem of task offloading between terminal devices and edge servers,a management and control framework based on SINET is designed,and a performance evaluation model for offloading strategies is proposed.First,the task offloading problem is described as a multi-queue system.Then,two offloading strategies(local first strategy and probability-based strategy)are analyzed using the Markov chain,and the mean response time of tasks and the average energy consumption of the system are derived.On this basis,an optimization-based resource adaptation method is constructed,considering device requirements,computing resources,and transmission resources.Finally,the experimental results show that the proposed model can accurately reflect the performance of the offloading strategies.The proposed resource adaptation method can dynamically allocate resources to satisfy the requirements of different devices(2)Focusing on the problem of the transmission resources collaboration with different edge networks,a functional model of transmission components based on SINET is designed,and a transmission scheduling mechanism for edge resources collaboration based on stochastic optimization is proposed.First,the transmission scheduling problem based on resource collaboration is formulated as a queue scheduling model,and a stochastic optimization problem is constructed.Then,the above problems are transformed and decomposed,and a low-complexity control algorithm is proposed,which can provide state-based transmission strategies.Finally,simulation results show that without loss of throughput,the proposed algorithm can effectively reduce the queuing delay of packets.(3)Focusing on the problem of storage resource cooperation between edge networks,a functional model of the cache component based on SINET is designed,and a collaborative caching approach based on multi-agent reinforcement learning is proposed.First,the optimization problem of cooperative edge caching is formulated,which is shown to be NP-hard.Then,a cooperative edge caching framework is proposed by combining the proposed SINET component and the reinforcement learning scheme.On this basis,a cooperative caching algorithm based on multi-agent reinforcement learning is proposed.The output of the proposed algorithm is defined as low-complex caching policies,with the purpose of reducing the complexity of the learning model while providing certain performance guarantees for each caching server.Finally,simulation results show that the proposed algorithm can improve the overall hit rate of the system by sacrificing the hit rate of individual servers.(4)Focusing on the problem of task offloading and service resource adaptation between the edge server and cloud server,a resource adaptation framework based on SINET is designed,and a resource adaptation mechanism based on service function switching is proposed.First,a stochastic optimization model is constructed to describe the resource adaptation problem,aiming at maximizing the benefits of task processing and minimizing the cost of service switching.Then,task requirements can be associated with service resources by exploiting virtual queue technology.After that,a low-complex algorithm is proposed.On this basis,an active task rejection mechanism is introduced to improve the performance of the proposed algorithm,providing deterministic queuing delay guarantees.Finally,simulation results show that the proposed algorithm is effective in reducing the queuing delay and can provide customized resource adaptation strategies.