Study Of Traffic Matrix Estimation In IP Network And Utility Maximization In Wireless Networks

With the exponential increase in the size of the IP network, the traffic matrix estimation is an important index of network design and measuring. Traffic matrix (TM) estimation problem is about the traffic estimation between all possible pairs of sources and destinations in the network, and it will provide data support and determination basis for network topology structure design, link measurements and routing information. IP network traffic matrix estimation problem has attracted more and more attention.Unfortunately, the direct measurement approaches for traffic matrix are generally unavailable because of the technique restriction, now the main approaches are inference approaches to estimate the traffic matrix. Generally, the relationship between the traffic matrix, the routing matrix and the link counts can be described by a linear equation Y=AX,So the research efforts in the area of the traffic matrix estimation have focused on the statistical inference problem about the traffic volume between OD pairs from Y and A. In IP network, especially IP backbone networks, the amount of the OD pairs are more than the link counts and it will lead to the several solutions for equation Y=AX So the challenge lies in the traffic matrix estimation problem is its high ill-posed nature.To overcome the high ill-posed nature, the part one in this thesis describes the inference problem with an Euclidean optimization problem, with the objective is to minimize the Euclidean distance between a certain predetermined prior and the target TM. Because the estimation solution is sensitive to the prior, based on the disadvantage of the existing methods generate the prior, we improve the Gaussian distribution of to generate the prior.Based on the Euclidean optimization model, we propose three novel methods for the traffic matrix estimation problem.(1) Solve the traffic matrix estimation problem by singular value decomposition and optimization theory. It will overcome the ill-posed problem by singular value decomposition and matrix transformations for the routing matrix then get the extreme value of the optimization problem as the solution for the traffic matrix estimation.(2) Solve the traffic matrix estimation problem by general matrix inverse and covariance matrix. It will overcome the ill-posed problem by the general solution expression by general matrix inverse and introduce the covariance matrix to capture the time-varying. Then by iterative process, we can attain the estimation of traffic matrix. This method gives the steps and method for the traffic matrix estimation on-line.(3) Propose a new method to realize the partitioning and the transformations of the routing matrix based on MPLM method, and give the new steps for the traffic matrix estimation on-line.Through both theoretical analysis and simulation results, it is shown that three methods achieve better performance than the existing representative methods and more accurate results compared with network actual traffic volume.The part two in this thesis have focused on the utility maximization in Wireless Networks.Based on the network utility and lifetime maximization in Wireless Sensor Networks, we study the trade-off problem between them, and propose a new method to solve the trade-off problem.A lexicographic method is adopted to solve this problem. First, we recursively induce the Max-Min lifetime for the cluster header. Then, under the given lifetime, we formulate the maximizing Utility problem as an optimization problem, and advocate the use of lexicographic method to achieve the optimal rate allocation solution. Theory analysis and simulation results validate and illustrate the effectiveness of the algorithm.Finally, the last part summarizes the dissertation, reviews the above research works and presents the future research directions.