Investigations On Network Virtualization Techniques For LTE

The development of society and economy has led to a huge explosion of the demands for wireless data as well as the requirements for diverse services.In future wireless networks,multiple technologies will coexist in a service-oriented infrastructure.Suffering from low resource utilization,vertical architecture,simplex network characteristic,solidification of network functions in the underlying hardware,etc.,traditional wireless network cannot undertake the development of future wireless networks.Wireless network virtualization(WNV)enables multiple virtual networks to share the same physical network,thus improves the resource utilization and reduces the capital expenditures significantly,and service providers(SPs)can support different kinds of services by configuring these virtual networks.As a result.WNV offers a new method for the coexistence of multiple networks and the deployment of new technologies.As the mainstream of the next generation of mobile communication,the virtualization for Long Term Evolution(LTE)is being widely studied.Wireless resource allocation in LTE network virtualization is investigated in this thesis.Firstly,LTE network virtualization is briefly introduced.The performance requirements,network architecture,key technologies and frame structure of LTE system are discussed.WNV is introduced and a general downlink framework for LTE virtualization is given.Wireless resource allocation in LTE network virtualization,game theory and its branch in economics,auction theory and applications of game theory in resource allocation in LTE network virtualization are introduced.For the case with a single infrastructure provider(InP)and a single evolved Node B(eNB),a virtualization framework and a Vickrey-Clarke-Groves(VCG)based resource block(RB)allocation scheme are proposed.In the framework,the hypervisor cannot access user information directly.RBs are first allocated from InP to mobile virtual network operators(MVNOs)and then MVNOs to their users.The allocation process of RBs from InP to MVNOs is modeled as a VCG auction.During each allocation round,each MVNO reports its valuations for RB sets to InP.InP does the allocation,aiming at maximizing the reporting value sum of all the MVNOs.The truthful revelation of VCG auction is utilized to encourage MVNOs to report true valuations.A unified utility function with configurable parameters is introduced,and MVNOs are able to configure these parameters to customize the scheduling strategies for their virtual networks according to their service requirements.A minimum RB demand is set for each MVNO to realize the isolation between virtual networks.Due to the high complexity of the RB allocation problem,a heuristic algorithm is proposed,where RBs are allocated one by one.During the allocation of an RB,each MVNO reports its valuation for the RB and InP allocates the RB to the MVNO with the highest reporting value.Two kinds of traffic,i.e.,traffic with quality of service(QoS)requirements and best effort(BE)traffic are considered.RBs are first allocated to QoS users and then among all users.Simulation results show that in the proposed RB allocation scheme,intra-slice customization and inter-slice isolation of virtual networks are guranteed.A virtualization framework for the case with multiple InPs and multiple eNBs is proposed.Under this framework,a game theory based rate request and allocation scheme is investigated.In the proposed framework,each InP owns a physical eNB.Each MVNO operates a virtual eNB under each InP.Virtual eNBs of the same MVNO are managed by a centralized controller owned by the MVNO.The procedure of rate request and allocation between InP and MVNOs are modeled as a two-stages dynamic game.In the first stage,each InP delivers the prices of rates to all MVNOs.Then in the second stage,each MVNO determines the rates required from each InP according to the prices from all InPs and its rate demands.In the network of each InP,an upper limit of requiring rate is set for each MVNO to guarantee the isolation between virtual networks in the same physical eNB.Since fixed upper limits are adopted,the circumstance may appear where in the same InP MVNOs with high rate demands cannot get more rates because they have reached their upper limits,while those with low rate demands have not reached their maximum requiring rates and there exist rates not required in their virtual networks.An upper limit adjustment algorithm is proposed to avoid this situation.The algorithm reduces the upper limits of MVNOs who have not reached their upper limits and increases the upper limits of those who have reached their maximum requiring rates.Then all players play the game with the adjusted upper limits.Simulation results show that,the requiring rate of a MVNO in an InP increases with the increase of the rate demand of this MVNO in this InP,but is limited by the upper limit.Thus the isolation between virtual networks in the same InP is realized.Meanwhile,the adjustment of upper limits enables MVNOs who have reached upper limits before to achieve higher rates while maintaining the inter-slice isolation of virtual networks,and gives the InP who does the adjustment an opportunity to gain more profits.