Researches On Customized Virtual Wireless Networks

Next generation mobile networks(5G)need to meet performance requirements for high throughput,low latency,extremely high reliability,huge connectivety,and high mobility.Moreover,in order to meet the needs of new vertical industries and application scenarios,the 5G network architecture uses software-defined network(SDN)and network function virtualization(NFV)technologies to transform the legacy mobile networks into a flexible and programmable structure.To this end,network slicing has been identified as a key enabling technology to implement flexible 5G architecture,which ensures the operation of customized logical networks running on shared infrastructure of radio access network(RAN),transport network and core network(CN).To efficiently deploy and manage end-to-end network slices,the following key issues should be addressed: network slice deployment,resource management among network slices,control plane configuration under coexistence of multiple network slices,network slice performance monitoring and reconfiguration.This dissertationis dedicated to the key technologies for realizing end-to-end mobile network slicing.According to the design principle of "the data plane functions of RAN slices are dedicated,and the CN control plane functions of network slices are shared",we analyzes the requirements of RAN slicing and CN slicing from the perspective of network,so as to solve the key technical problems of customized wireless virtual network.The contributions and innovations of this dissertation are as follows:According to the requirements of “customizable” and “elastic” of RAN slices,we address the deployment issue of the data plane of RAN slices.Since the data plane function of RAN slice needs to be quantified in an on-demand manner according to different use cases,we propose software defined protocol(SDP)technology,which is based on the central control management and infrastructure virtualization.SDP allows the protocol function blocks to be tailored to facilitate a flexible service-oriented protocol stack deployment for the data plane of RAN slices.Furthermore,in order to implement the SDP technology in data plane slicing for the RAN layer 2,we address the fundamental issue of the software defined protocol stack mapping(SDPM).SDPM deals with data plane request mapping for a RAN slice with specific Quality-of Service(QoS)onto the underlying network through executing our proposed decomposition algorithm to select the optimal SDP function blocks and virtual links.Finally,the implementation framework of SDP technology is provided,and the experimental results show that SDP technology can effectively realize the deployment the data plane of RAN slices with satisfactory performance of stack processing delay,mapping cost and resource utilization.For the issue of "end-to-end slice guarantee",we solve the problem of robustness of SDN/NFV technology based RAN slice and the reconfiguration of the slice triggered by the user’s uncertain service demands.First,the network slice recovery problem is solved by re-mapping a network slice with deterministic requirements,where remapping is the process of reselecting the virtual network functions and links to accommodate the failed network slices.The solution to this problem is used as a benchmark to validate the performance of robust network slicing algorithms.Second,two robust network slicing algorithms are proposed for the recovery and reconfiguration of network slices under stochastic demand.Finally,we run all the proposed algorithms on the testbed of the data plane of RAN layer 2.Numerical results reveal that our proposed robust network slicing algorithms can provide adjustable tolerance of traffic uncertainties.As a supplementary to research on RAN slicing,the sharing of key functions of multiple slices on the CN is investigated with the aim of realizing end-to-end slicing and reducing signaling overhead.In particular,we consider the problem minimizing the CN signaling overhead while multiple slices share mobility management functions in the CN.More specifically,to meet the requirement of "key function sharing among slices",we study the problem of configuring the optimal paging and location tracking area of the user in a slice.Based on the multi-level tracking area and the multi-level tracking area list structure,we construct an embedded Markov chain model to analyze the relationship between the location update frequency and the paging frequency of a specific UE in the network slice and the resulted theoretically optimal tracking area lists are allocated to UEs per each slice.By comparing the numerical and simulation results,the effectiveness of our proposed analysis model is validated.Through applying the proposed multi-level tracking area structure,we can reduce the average of paging and location tracking signaling overhead in comparison with the legacy single level tracking area strategy.For the requirement of “multiple slices deployment”,we deal with the load-balancing problem among CN mobility management entities(MMEs)in the virtualized resource pool.With NFV technology,the 5G supports the virtualization of MMEs and MME VNF pool management in CN.Therefore,considered that multiple network slices coexist in the network requiring to share the MME paging resource,we need to address to the problem of allocating paging resource in a load balanced manner among MMEs in the MME Pools.First,the signaling cost of a single network slice at its CN’s control plane is minimized,including paging and location area tracking.Second,considering that the UE can be configured at different state models(i.e.,RRC_idle or RRC_Inactive),we design a procedure for paging the UEs of different slices and for UEs performing tracking area update.Finally,we formulate and solve the paging resource load balancing optimization problem among MMEs residing in different MME Pools when multiple slices coexist.By comparing the simulation results of the proposed scheme with the benchmark algorithm,we can effectively distribute the paging load among MMEs in the MME Pools under the coexistence of multiple slices scenario.