Research On Utility Based Network Resource Allocation

Due to the poor quality of service (QoS) support of traditional network resourceallocation methods, in this dissertation, utility-based network resource allocationmethods were studied. This kind of approaches stand at the view of the user and becomea hot point in research community recently. Although many research achievements hadbeen made, there were still a lot of challenges, such as the rough resource allocationgranularity, the poor fairness, the complex resource allocation optimization model andthe inefficient model solution algorithm. This made it difficult to put these methods inuse in real network devices. According to the above mentioned facts, novelty utilitybased methods were proposed for various application scenarios, which are described asfollows.Aiming to the fine granularity resource allocation, a two layers scheduling modelnamed U-S (the Utility Class and Service) was proposed. First, the curves of the utilityfunction clusters were introduced and the corresponding mathematical formulas weregiven, which could distinguish different applications in aggregate traffic effectively andimproved the fairness among various applications. Secondly, for reducing thecomputational complexity of the model, an effective and proper model solutionalgorithm was employed. Based on the time complexity analysis, the accuracy analysisand the algorithm convergence analysis as well as the experimental verification, theproposed approaches showed high effectiveness and efficiency.For the purpose of improving the universality of the resource allocation model andthe running speed of the solution method, a marginal utility function based networkresource allocation model was proposed. In first place, the network applications wereclassified according to the features of their marginal utility. And then the marginalutility functions of each traffic category were given and used to derive thecorresponding utility functions which were the basic components of the utility functionframework. Compared with the traditional utility function generating methods, theabove-mentioned approach is more universal and can simplify the resource allocationmodel. After the marginal utility based optimization model was constructed, thenecessary conditions for the optimal solution was proposed and the model solving algorithm named MU (Marginal Utility) was described with the accuracy analysis andthe convergence test. Simulation experiments show that compared with the state-of-artapproaches, our method has higher effectiveness and faster solution speed.To address the issue of resource allocation in wireless network, we proposed atwo-stage utility based wireless network resource allocation method. Most utility basedwireless resource allocation methods only focus on the optimal allocation within a basestation. Because of ignoring the issue of base station or network selection, thesemethods can not achieve the optimal utility of the whole network. From our observation,we could notice that the base station selection was the key point for a user to obtain theoverall optimal utility in the overlap area of multi-base-stations, as well as the basementseamless switchover or roaming in the future heterogeneous wireless network.Therefore, the process of wireless network resource allocation can be divided into twostages. In the first stage, base stations are selected according to their congestion degree.And in the second stage, based on the characteristics of the wireless network, a wirelessnetwork resource allocation algorithm derived from MU is employed. The simulationexperiments show the rationality of the base station selection method as well as theefficiency and effectiveness of the resource allocation algorithm.