| The development of the Internet brings convenience to people's life,but also brings great challenges to network performance monitoring.Network measurement is mainly measured network environment performance indicators,including delay,packet loss,and bandwidth and so on.Distributed Network Measurement System is a distributed measurement system designed to measure the performance of a given area network.By deploying measurement probes in network nodes to perform measurement tasks,performance monitoring of the network under test can be achieved.In a small-scale network environment,using all network nodes as measurement probes to deliver measurement tasks can achieve full-link monitoring of the network to be measured and evaluate the performance of the network more comprehensively;in a large-scale network environment,traffic loss,hardware and software resource consumption caused by full-link monitoring cannot be ignored.Too many measurement tasks may even cause blocking of the network under test.Therefore,how to reduce the resource occupation of the network to be measured by the measurement task is an important issue in network measurement.In order to solve this kind of problem,we can optimize it from two aspects:one is to improve the measurement algorithm,reduce the hardware and software resource consumption and traffic load caused by the measurement tasks,and the other is to optimize the measurement node selection scheme.Using fewer nodes as probes to measure performance across the entire network.This paper aims to propose a measurement node selection scheme,by deploying measurement probes in the key nodes in the network to be measured and collecting network performance indicators,to achieve the purpose of reducing resource consumption and network load.The optimization objective selected by the network measurement node is to collect the network performance indicators by performing measurement tasks on the nodes with less load and achieve the purpose of monitoring the performance of the entire network.Therefore,the selection problem of network measurement nodes is a combinatorial optimization problem considering the average load of nodes,the number of nodes and the network coverage.In this paper,by using the heuristic algorithm of ant colony algorithm and genetic algorithm,combined with the measurement node selected application scenarios,the design and implementation of the measurement node automatic selection algorithm is proposed,and add a variety of local optimal solutions of the basic ant colony algorithm and genetic algorithm to evasion strategy,and optimize the search ability of the algorithm for the global optimal solution.In order to improve the execution efficiency of the selected point algorithm,this paper designs and implements the parallelization framework of the point selection algorithm based on the message transfer model.Experiments show selection results of the algorithm based on ant colony and genetic algorithm are superior to the traditional algorithm in terms of the average load of nodes and the number of nodes.When the parallelized point selection algorithm is used to deal with the problem of large-scale network selection,the efficiency of the algorithm has been significantly improved compared with the non-parallel algorithm.In this paper,we first analyze the mathematical model of the node selection problem and the commonly used algorithms at the present stage,and then design and optimize the basic ant colony algorithm and the main modules of the genetic algorithm by combining with the scene of the selection point,and realize the selection based on ant colony algorithm and genetic algorithm.Then a series of frameworks for parallelizing point selection algorithm are designed and implemented by studying a variety of parallelization schemes.Finally,we propose measurement node selection experiment and extremal function of problem solving experiment to compare the efficiency of different node selection algorithm. |
PDF Full Text Request