Research On Network Performance Detection Technology Based On Direct And Indirect Model

With the continuous expansion of network scale and the applications of various complex services, how to effectively obtain network performance data has become particularly important to network management and improve network quality of service. At present, the network detection is one of the most direct and effective means to obtain network performance data through. Depending on the detection way, the network detection methods are generally divided into direct detection and indirect detection. Direct detection method through setting auxiliary detection point on the nodes or links, record and statistic the network flow information. Indirect detection through sending probe packets to the target detecting link or node to get network performance, such as: end to end delay, packet loss rate, bandwidth and so on. However, whether the direct detection or indirect detection, detection cost is inevitable. It consists of two parts, the detection point of deployment cost and the impact cost raised by detection act itself. Network detection cost directly determines the efficiency and accuracy of the network detection. Therefore, the network detection technology research focuses on how to effectively reduce the cost of network detection.In indirect detection model, mainly through reducing the number of detection point deployment and probe packets to reduce the cost of network detection. For the optimization problem of detection points is usually converted to the minimum vertex cover problem. If given detection points, generally through optimizing the detection assignment scheme to reduce the cost of detection; In direct detection model, optimization method is to select fewer points detection and minimize duplication detection of network node or link as much as possible. By using the flow conservation law, the cost optimization problem can be abstracted as weak vertex cover problem of the undirected graph, thereby reducing the number of detection points and achieving lower detection cost.In this paper, we focus on optimizing the detection assignment scheme of the indirect detection model and the detection point selection problem of the indirect detection. For indirect detection, by considering the links use to reduce the number of detection probe packets, in this way can be based on the existing detection result, to dynamically adjust the assignment of probe packets between the link and node, thereby reducing the impact on network performance with detection process; For direct detection, based on the flow conservation, the problem of detection point selection can be converted to weak vertex cover problem of undirected graph, and using the greedy algorithm to solve the weak vertex cover set, and then remove redundancy nodes by backtracking to the result set. In order to verify the effectiveness of the proposed algorithm, we simulated the algorithms above, through comparison with other algorithms to prove that the algorithm which proposed in this paper can get better detection point set and smaller detection cost.