Research On Resource Allocation Optimization In Space Information Networks

With the continuous development of computer technology and space technology,the space information network has attracted attention of many countries,and has become a hot spot of current research.As there are too many nodes in space information network and they come from different sources,the control of every node and collaboration between different nodes are difficult.Due to the uneven distribution of ground users and the unpredictable demand of users' needs,it may cause network congestion or service interruption when providing information services for ground users.Therefore,how to guarantee the service ability and quality of the network under the limited resources has become a significant research work in the space information network.After a full analysis of the space information network and network traffic offloading problems,this paper proposes the near earth orbit satellite traffic offloading model based on software defined network architecture.To solve the model,this paper designs distributed algorithms based on the dual decomposition algorithm and alternating direction method of multipliers.Also,delay and packet loss problem in space information network are discussed in the end.The main research contents and innovations of this paper include:(1)To solve the problem of uneven distribution of resources in space information network when providing data service for ground users,we introduce software defined network architecture and make GEO satellite the SDN controller to control the whole network.Then we propose the near earth orbit satellite traffic offloading model.To evaluate the model,we calculate the amount of offloading data and the overall network utility for networks in different scale.The simulation results show that the model can effectively improve the resource utilization and service capacity of the space information network.(2)For the near earth orbit satellite traffic offloading model,we propose the Traffic Offloading algorithm based on Dual Decomposition(TODD)to solve the model distributedly.Then we point out that TODD has defects in privacy and convergence speed.To accelerate convergence speed,we propose the Momentum enhanced Traffic Offloading algorithm based on Dual Decomposition(M-TODD)based on the idea of momentum.The simulation results show that the proposed algorithms both work well and the convergence speed of M-TODD is better.(3)For the near earth orbit satellite traffic offloading model,we propose the traffic offloading algorithm based on Gauss-seidel ADMM(G-ADMM)to solve the model distributedly.Then we point out the advantages of the algorithm in the utilization of computing resources and privacy.Considering that multiple modules may not compute parallelly in the Gauss-Seidel ADMM,we propose the traffic offloading algorithm based on proximal Jacobian ADMM(J-ADMM).In J-ADMM,every node is computing node,so as to make full use of computing resources in the network and ensure the efficiency of algorithm.The simulation results show that the proposed algorithms both work well and the convergence speed of J-ADMM is slower.Theoretical analysis shows that this is the effect of proximal term in J-ADMM.(4)For delay and packet loss problem in space information networks,we propose two kinds of variable updating strategy(KU and PMT)to adjust the network traffic offloading process.Then we combine the two strategies with G-ADMM and J-ADMM to generate algorithms for networks with delay and packet loss(KU-G-ADMM,KU-J-ADMM,PMT-G-ADMM and PMT-J-ADMM).Simulation results show that,for space information networks with delay and packet loss,the momentum based prediction strategy(PMT)guarantee the convergence speed and stability of the algorithm effectively.