C-RAN Network Load Prediction And Load Balancing Using Sparse Matrices

C-RAN architecture,characterized by centralization,collaboration,cloudization,and cleanliness,is considered to be an important development direction for future wireless access networks.In the future,mobile communication networks have high node density and large amount of data,resulting in high complexity of load prediction and load balancing.The sparse matrix can reduce the computational complexity due to the large number of zero elements,and it has become an essential key technology in the field of numerical computation in science and engineering.Therefore,it is important to apply the sparse matrix technology to load prediction and load balancing of wireless networks.Aiming at the problem of high complexity for solving linear equations caused by the existing C-RAN architecture based on the Markov load prediction model without using the sparsity of large-scale network state transition matrix,it is proposed to perform block iteration on the network state transition matrix,divides 4 blocks each iteration,and defines the offset of 4 matrices respectively.When one of the offsets of the block matrix belonging to the same column is tested as a zero matrix,it is directly derived that all the elements of the corresponding block of the requested matrix are zero,then proceed to the next iteration;When none of the offsets of the block matrices belonging to the same column have been checked out to be zero matrices,we convert into an iterative format by transfo rming the matrix equations,and then processed in blocks,thus reducing the complexity of the solution.The critical value of sparse matrix sparsity is quantitatively analyzed by simulation,and the relationship between the sparsity and the computation is given and its rationality is proved.The correction scheme of the load transfer matrix is designed for the dynamic change of the external environment.The simulation results show that the algorithm can reduce the complexity of the solution of the load transfer matrix without affecting the prediction accuracy.Aiming at the problem that a load balancing mechanism based on markov load prediction for C-RAN architecture,the existing algorithm adopts an iterative method to adjust the power,However,when the network scale is large,result in high complexity,a load balancing mechanism using sparse matrix prediction is proposed.The mechanism,aimed at minimizing the load transfer correlation degree and minimizing the load transfer balance degree,and combining the Ncut cutting algorithm in the graph theory to obtain the best sparse matrix block,achieves the purpose of dimension reduction and zero division of the load transfer matrix,after the block,the load transfer matrix of each block is recalculated,and the load transfer matrix is used for load prediction.Finally,the power adjustment step size is determined according to the predicted load,and the pilot signal power of the cells in each block is adjusted in parallel to achieve load balancing.The simulation results show that the load balancing mechanism proposed in this thesis can reduce the complexity of load balancing.