Abstract
In a distributed architecture, a router consists of multiple independent line cards each taking forwarding decisions based on the locally stored forwarding Information base (FIB). FIBs on the line cards must store the enough prefixes to take correct forwarding decisions. In a default free zone, the continuously growing size of Internet is increasing convergence time, increasing the cost of the routers, and reducing the forwarding speed. This paper proposes a simple and new idea for reducing the amount of network prefixes stored in Internet routers line cards, thus reducing routers load and improve routing convergence without changing routing protocols semantics. After the routing protocol calculates the path to a prefix, the routing information base (RIB) has the outgoing interface(s) and the next-hop(s) pointing to the downstream router(s). The idea is for upstream routers to inform its downstream routers about the prefixes that the upstream router is using as a next-hop. On the downstream router, the line card facing the upstream router needs to store these prefixes only. Other line cards need not store these prefixes. The proposed algorithm applies to any routing protocol. However we choose BGP in our experiments because BGP is the biggest generator of prefixes among all routing protocols. By analyzing real BGP data obtained from router views, we show that the approach results in a significant reduction in the number of prefixes stored on line cards. As an application to the algorithm, we propose a new option in BGP ORF message for an upstream router to inform its downstream routers about prefixes.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have