IPv4 and IPv6 Troubleshooting Enhancement through Reverse Path Discovery

摘要

The analysis of a network path between a source and a destination using traditional networking tools like ping and traceroute, does not provide informations about the return path. In the absence of those informations, malfunctions and faults can not be managed with adequate effectiveness. In this paper we propose two techniques to perform network path discovery in production networks, without the necessity of a centralized network monitoring system. The first solution is suitable for IPv6 networks, while the second one is suitable in both IPv4 and IPv6 contexts. In order to pursue troubleshooting efficiency and because of possible restrictive policies, the path discovery is triggered by acting only on a single device (from now on called the Discovery Router), without the need of retrieving informations on multiple network equipments. The IPv6-only approach relies on the use of the Hop-by-Hop Option Extension Header without the definition of new options; the second proposed approach aims at the activation on the target device of a traceroute back through the discovery router, then combining the capabilities of SNMP and TFTP. Therefore, it is possible to identify the path between any two routers, for both directions, in a traceroute-like format. A key important issue for telecom management is the delay in comprehending network conditions. This delay harms precise network management activities because the amount of managed data required for performance management has been increasing due to emerging technology such as the virtualization tailoring dedicated network service for an enterprise. Hence, the increase in managed data is accelerating, and in future, we will see a fatal limitation of the current management approach. Timely monitoring of failure is increasingly valuable. To tackle this issue, a decentralized performance management system is assumed, and a real-time performance monitoring implemented inside distributed network devices is proposed. In the system the demonstration result shows that burst traffic was successfully detected at 250 millisecond intervals, and a detection resolution of its failure is significantly improved, compared with the conventional centralized management approach.