A scalable on-line multilevel distributed network fault detection/monitoring system based on the SNMP protocol

Traditional centralized network management solutions do not scale to present-day large-scale computer/communication networks. Decentralization/distributed solutions can solve some of these problems (Goldszmidt, G. and Yemini, Y., 1995), and thus there is considerable interest in distributed/decentralized network management applications. We present the design and evaluation of an SNMP-based distributed network fault detection/monitoring system. We integrate into the SNMP framework our ML-ADSD algorithm (Su, M.-S. et al., Proc. 39th Annual Allerton Conf. on Commun., Control, and Computers, 2001; Su, "Multilevel distributed diagnosis and the design of a distributed network fault detection system based on the SNMP protocol", Ph.D. Thesis, School of Computer Science, University of Oklahoma, 2002) for fault diagnosis in a distributed processor system. The algorithm uses the multilevel paradigm and requires only minor modifications to be scalable to networks of varying sizes. The system is fault tolerant, allowing processor failure and/or recovery during the diagnosis process. We have implemented the system on an Ethernet network of 32 machines. Our results show that the diagnosis latency (or time to termination) is much better than that of earlier solutions. Also, the system's bandwidth utilization is insignificant, demonstrating the practicality of its deployment in a real network. We have successfully integrated three modern disciplines: network management, distributed computing and system level diagnosis.