An intelligent self-learning algorithm for IP network topology discovery

The significance of network topology discovery cannot be denied, especially for tasks like network management, network analysis or network visualization. In this paper we describe a novel topology discovery algorithm which is intelligent, efficient and self-learning. Sending ICMP requests to inactive hosts can waste considerable amount of time in the discovery process. We propose an algorithm that queries hosts having higher probability of being active. Our algorithm is self-learning in the sense that it can learn and decide for itself which ranges of IP addresses to send ICMP echo requests that would yield quick initial response. Our algorithm does not entirely rely on SNMP-MIB or ICMP echo request/reply, DNS, Trace route etc, rather SNMP is installed only on routers, switches and network printers. We have implemented and tested the algorithm at NUST Institute of Information Technology, Pakistan and it has accurately discovered the network topology