Research On Function Orchestration Technology In Reconfigurable Network Based On Atomic Capability

Currently,the rigid IP architecture and best-efforts service model of the Internet can not satisfy stringent requirements of the diverse service characteristics,ubiquitous service deployments,integration of heterogeneous networks and customized function provision.The big gap between service requests and internal network capability requires novel network architecture.Therefore,Flexible Architecture of Reconfigurable Infrastructure(FARI)constructs a basic physical network that can dynamically reconfigure and develop functions through the enhancement of network layer and transport layer features,which can provide customized basic network services for different services.It greatly improves the adaptability,evolution and extensible ability of network structure for characteristics and requirements of evolutional services.In terms of network function orchestration,in order to solve the problem of bloated,inefficient and complicated traditional network resulted from increasing the network basis service capabilities with “patching”,Flexible Architecture of Reconfigurable Infrastructure abstracts and models the network function of the core layer as more fine-grained atomic capabilities,combines the network cognition and network reconfiguration technology,constitutes the basic service unit by mapping atomic capability to atomic service at the node level,and constructs service chains by mapping atomic service to service of the network level.Finally,it completes the dynamic multi-level orchestration of the networking functions.Supported by the Research on Flexible Architecture of Reconfigurable Infrastructure,based on the atomic capability and Reconfigurable technology,this dissertation researches on dynamical clustering of atomic capability instances,placement on-demand of function instances,flexible construction and reconfiguration of service chains.We achieve the dynamic adaptation between network status and service requirement.The main contributions of this dissertation are as follows:1.Performance compatibility between different atomic capability instances caused by fine-grain function decomposition may cause the performance reduction of atomic service in FAIR.In this dissertation,we propose the performance compatibility based atomic capability instances clustering method,and realize the accurate clusters of atomic capability instances.First,we define the performance compatibility among some atomic capability instances as the fitting between,and formulate the hypergraph clustering model that is high-order and multidimensional and therefore difficult to solve.Second,we cast it into an evolutionary game problem and prove that the Evolutionary Stable Strategy of evolutionary game is the solution of hypergraph clustering exactly.Finally,we design the replicator dynamics equation to solve the problem.The experimental results demonstrate that the proposed clustering method can keep error ratio less than 5%,and improve the performance of atomic capability combination.2.It needs to optimize the placement and resource allocation of function instances according to their dynamic load.So we propose a placement strategy of function instances that maximizes resource utility,ensuring the quality of service with the least redundant resource.First,we analyze the two scenarios of the dynamic placement of the function instances,and propose the number decision and location decision of function instances.Second,based on the online load forecast and load rate double thresholds mechanism,we formulate the node resource occupancy rate minimization model and solve it with simulated annealing algorithm,make the optimal number decision for function instances placement.Finally,the resource utility maximization model is established and solved by the penalty function based hybrid particle swarm optimization algorithm for the updated function instances,and the optimal location decision is realized.The simulation results show that the two proposed strategy can maximize the effectiveness of network resources and guarantee the quality of service simultaneously.3.The mapping between atomic service and service at the level of network needs cooperative control of the whole atomic service instances.Therefore,the load balance oriented service chain deploying framework and method is proposed,achieving the flexible adjustment of network function and load balance simultaneously through selecting the optimal atomic service instances for the service chain.First,the deployment framework of service chain and its core module deployment engine are designed,that can select and link the optimal atomic service instances for the service chain.Second,the deployment of service chain is formulated as an integer linear programing model whose objective is to maximize utility that is evaluated by the degree of load balance of the atomic service instances,and the contribution degree based node resource priority algorithm is designed to solve the model.Finally,we implement the deployment engine atop the OpenDaylight controller and conduct experiment to prove that the proposed method can satisfy the different functional and performance requirements of the services,while ensuring the network load balance.4.For the reliability problem of service chain caused by node resource bottleneck,we propose the service chain reconfiguration scheme based on migrating heavy load atomic service between nodes to achieve the quick reconfiguration of service chain with minimum cost.First,based on the profit of flow,resource of destination node,quantity of state information and update of flow table,we establish the reconfiguration utility maximization model for the service chain reconfiguration problem.Second,we introduce a parallel reconfiguration algorithm based on the Markov chain,to allow each reconstruction thread to calculate independently and concurrently.Finally,we design the interface of north and south for the controller and the atomic service migration scheme that supports the consistency of the flow state.The simulation results show that the proposed method can improve the service chain reconfiguration effectiveness and ensure the service quality of the service chain.