Routing State Based Centralized Network Fault Detection And Repair System

Over the past decade,the Internet has achieved great success,has become an indispensable part of daily life.As network size increases and new network devices continue to be applied,network management becomes more complex.According to the relevant research,the frequency and the number of network failures also increased year by year,causing incalculable losses to people.Therefore,how to automate the detection of network faults and repair network failures has become one of the current problems of network management.At present,the existing network fault handling mechanism can automatically detect network failures,but it requires network administrators to manually repair the network failure.This mechanism is difficult for network fault detection and repair to provide efficient solutions,mainly in the following questions: 1)Network fault repair lacks of intelligence,so it can't reduce the impact of network failure on the network in the shortest possible time;2)Existing network fault models deal with complex types of faults,increasing the complexity of network fault detection systems;At the same time,part of the fault type is limited by the hardware platform,so the upgrade and maintenance of the system is inconvenience.To solve the above problems,this paper proposes a routing state based centralized network fault detection and repair system,namely RS-NDRS.It fully adopts the idea of I2RS(Interface to Routing System),which opens control protocol of the routing interface to the upper application,so that applications can use routing state information to detect network failures.Compared with the traditional network detection mechanism,the routing state information is relatively small and the data type is fixed,so RS-NDRS can detect the network fault efficiently.At the same time,in order to avoid the impact of network failure on the network,it can take the initiative to issue a security strategy to the network node to suppress the impact of network failure.In order to efficiently obtain the network state information of the network nodes,two mechanisms are proposed,which are adaptive polling topology aware mechanism and adaptive polling period adjusting mechanism.The adaptive polling topology-aware mechanism is a mechanism to solve the problem of how to dynamically adjust the monitoring range according to the change of the detected equipment.The experimental result shows that the algorithm is better than the traditional polling-based mechanism for network topology changes with a better response ratio.Based on the analysis of the traditional polling and event notification mechanism,this paper proposes an adaptive polling period adjustment mechanism which dynamically adjusts the interval of collecting the routing state information according to the historical state information to reduce the load pressure on the network caused by the routing state information collection as much as possible.The experimental result shows that the algorithm can balance the fault response time and the polling overhead.Fault detection and repair algorithm based on production rule is adopted to detect and repair the network fault.This mechanism detects and automatically fixes known faults.And the mechanism provides early warning of unknown faults and further fault diagnosis by network management staff to ensure the accuracy of fault detection.And a novel fault-matching algorithm based on tree depth-first traversal is proposed.Finally,the RS-NDRS prototype system is built based on commercial Zeb OS software,and uses the next generation network management protocol NETCONF as the management protocol for routing state information acquisition and configuration.In order to further verify the feasibility of RS-NDRS,an experimental network consisting of 50 network nodes is built,and each network node is deployed with OSPF protocol.The experimental results show that the maximum network fault detection time is maximum of polling circle T2 minutes and the security policy delivery time is 0.3 seconds.Finally,by demonstrating the OSPF router-id collision detection,the system can effectively detect and repair the fault.