| Abstract: | There is evidence based on previous failures that link state protocols such as PNNI may not be able to recover from a widespread loss of topology database information or an overload of topology state updates. To prevent this, PNNI extensions have been adopted in the ATM forum to avoid going into congestion and to recover when congestion occurs. This article presents the adopted mechanisms to enable the PNNI routing protocol to avoid congestion states wherever possible, to respond gracefully to network overloads and failures, and to recover from massive loss of topology database information. The mechanisms include detecting congestion and notifying peers, reducing the rate of topology update control messages, maintaining link adjacencies with prioritized treatment of critical control messages, database backup, graceful recovery of lost topology database information, and database resynchronization. Simulation analysis demonstrates the effectiveness of these routing congestion control mechanisms. |
