Researches On Service-oriented Network Fault Management Technology

With increasingly complex and isomerized network environment, people more andmore rely on the network service to handle daily affairs. Once the network fault andcan’t be repaired in time, the loss may be significant or even catastrophic. Networkmanagement is increasing demanded, and the main problem it faces includes networkconfiguration, fault, performance, security and statistics. Network fault management isan important part of network management, targeting at quickly detecting and correctingnetwork failures and dynamic maintenance of network availability and reliability. Thisis of great significance for a stable and reliable computer network.Service-oriented network fault management is the key of this thesis. A hierarchicalnetwork fault management system based on the correlation graph and case library isdesigned. It receives and processes the message of managed network object based on aset of predefined management strategies to complete network fault monitoring, faultdiagnosis and fault processing function. Five modes including polling MIB, Trap,capturing ICMP packet, Ping and services testing are designed for the fault monitoringmodule in the system for monitoring the network object state. Before network faultdiagnosis, there is a great number of alarm events need to be filtered, removingredundant alarms and providing more streamlined information for fault diagnosismodule. In the alarm filtering module, a self-adjusted threshold algorithm is designedfor problems with extra large or small threshold. The threshold can be dynamicallyadjusted in this algorithm. A redundancy filtering algorithm is also designed in thisalgorithm, which play a significant role for repeating and oscillating alarm.In the fault diagnosis module, the origin and characterization of the network fault isstudied, and a hierarchical fault management model is constructed. There are four levelsin this model including the user interface layer, service interaction layer, platform andnetwork layer. The fault diagnosis process will start in the layer where the fault is found,and information would also be exchanged among different layers. Finally, a faultdiagnosis algorithm based on correlation graph and case library is improved for networkfailure diagnosis on each layer. This algorithm can express the logical relationship suchas "and" and "or", making use of graph properties to improve the efficiency of diagnosiswith the support of case library.In order to make the system has a good cross-platform characteristic and scalability, a SOAP communication module is designed. The popular network managementtechnology based on Web is adopted. The unified management interface makes thenetwork fault management more convenient and flexible without geographicalrestrictions.