Research On Active Network Fault Detection And Recovery Methods

With the development of economy and society,the scale of the network is getting bigger and bigger,and the network security management has gradually become an important part of modern network technology.Network failure is the focus of network management.The external manifestation of failure is network symptoms,which can be observed through different ways.At present,the commonly used observation methods are network alarm,active and passive observation mechanism,etc.The existing fault detection technologies generally set up a certain number of detection stations,send detection packets to the network to obtain the status of nodes or links in the network,and locate the fault links or fault nodes in the network.The nodes with high data transmission ability in the communication network have high real-time load and high transmission value.If only the probe number is considered when choosing the detection path,the probe in the large-scale network is longer,it is easy to cause long detection time and the detection packet loss,leading to high detection cost when fault detection.Moreover,in task-oriented networks,each resource-limited node needs to participate in multiple tasks.In the same task,multiple nodes cooperate with each other to undertake the assigned task.Each node needs to be able to complete multiple tasks simultaneously without failing with the maximum use of the resources.However,when faced with various attacks,the failure of core nodes is highly likely to lead to large-scale cascade failure,and even cause network failure.To address the above problems,this paper first proposes an active detection algorithm based on differential node weights to select the appropriate detection path to reduce the detection cost.The algorithm defines the node weights to measure the data transfer capability of the nodes.In the detection station selection stage,the algorithm iteratively selects the node with the smallest value as the detection station;in the detection path selection stage,the probe length is limited through the appropriate K value to reduce the probe round-trip time.To ensuring that all nodes in the network are detected,the algorithm selects probes that meet the conditions,and then expands the node coverage to reduce the number of probes and reduce the detection cost.Simulation results of stochastic and real network topologies show that the proposed fault detection algorithm can effectively reduce probe number and reduce detection cost compared with others.Secondly,considering the dynamic changes of the network,using the percolation theory,this paper proposes a fault recovery algorithm based on the network adaptive strategy to improve the detection efficiency.For the limited node resources,we reconfigure the network topology according to the unused resources of the node and the partial load of the node in each task group and the node load limit of each task group.The information transmission amount of the fault node and the traffic of the overload node beyond the node load limit are balanced to reduce the working pressure of the node,improve the network adaptability and enhance the fault tolerance of the network system.After a large number of simulation experiments,compared with the previous load average distribution algorithm,the proposed fault recovery algorithm based on capacity difference significantly improves the performance of node utilization,adaptability cost and overload node proportion,which improves the network adaptability and effectively prevents the occurrence of large-scale cascade failures.