Research On Load Forecasting Method And Migration Scheduing Strategy For Carrier Migration In C-RAN

C-RAN is a new type of green wireless access network architecture,which constructs a baseband resource pool through the centralized placement of processing resources,and abstracts baseband resource pool into virtual base stations based on virtualization and software defined radio technology to provide services for the communications users.This architecture can greatly saves the costs of base station' construction,maintenance,and expansion upgrading.The carrier migration mechanism in C-RAN is the key to implement resource sharing,hotspot eliminating and disaster tolerance in baseband resource pool.The essence of the carrier migration mechanism is the process of migrating the load service in the high-load virtual base station to other low-load virtual base stations in real-time non-perception.Firstly,the carrier migration mechanism needs to determine which virtual base station's load service needs to be migrated and where the load service should be migrated to by performing load state forecast on each virtual base station.Secondly,it is necessary to find an efficient migration scheduling scheme for the load service to be migrated so that the cost of the migration could be as low as possible.Therefore,how to forecast the load of the virtual base station and how to carry out the migration scheduling of the load service are two key issues in the C-RAN carrier migration mechanism.This paper mainly focuses on these two issues.The main work and innovative points of this paper are as follows:1.Concerining the problem that how to forecast the load of virtual base stations,this paper firstly analyses the change rule of the load services on the virtual base station and points out the two characteristics of tidal effect,periodic and abursty characteristic of self-similar and long-range dependent to describes the feasibility of the forecast for the load state of the virtual base station.After that this paper uses the GRNN to establish the load forecast model to excavate the regular characteristics of historical data.The particles would frequently enter the redundant space for searching when the classical PSO algorithm carried out the training of model,thus resulting in the number of iterations required to finish the training is too large.Aim at this problem,an improved PSO algorithm based on the interval eatimation of high quality solution is proposed to train the load forecast model.Finally,the simulation results show that the load forecasting method based on the improved PSO-GRNN can effectively predict the load state of the virtual base station.2.Concering the problem that how to carry out migration schedule for load service,firstly,this paper puts forward the strategy of scheduling migration which uses single load service as the scheduling unit to get rid of the shortcomings of the existing migration strategy,which are the time-consuming and the high cost.Moreover,a migration scheduling optimization model is estabilished based on the loss of total service value of the system.Since the established model is a NP-hard problem,the existing heuristic solution method for this problem can not be applied to carrier migration scene,which has a high demand for solving speed.This paper then proposes a two-stage approximation solution method.The first stage of this method groups the load service and determines the migration priorities of each group.The second stage uses the LR algorithm to solve the migration scheduling sub-model of each load service group to get the specific migration scheduling scheme.Meanwhile,an improved strategy of directional correction is proposed to solve the problem which is that the iterative updating direction of the multiplier (?) of the LR algorithm is easy to deviate from the optimal direction in the second stage.Finally,the simulation results verifiy the validity and superiority of the proposed two-stage carrier migration scheduling algorithm based on improved LR algorithm.