IP Routing

Summing Up

IGRP has the robustness of RIP but adds a major new feature -- route metrics based on bandwidth and delay. This feature -- along with the ease with which it can be configured and deployed -- has made IGRP tremendously popular for small^[3] to mid-sized networks. However, IGRP does not address several problems that also affect RIP:

• The exchange of full routing updates does not scale for large networks -- the overhead of generating and processing all routes in the AS can be high.

• IGRP convergence times can be too long.

• Subnet mask information is not exchanged in IGRP updates, so Variable Length Subnet Masks (VLSM) and discontiguous address spaces are not supported.

These issues may be too significant to overlook in large IP networks in which address-space conservation may necessitate VLSM, full route updates would be so large that they would consume significant network resources (serial links to branches tend to saturate quickly, and smaller routers may consume a lot of CPU power just to process all the routes at every update interval), and the convergence times may be too long because of the network diameter. Even small to mid-sized networks may choose not to implement IGRP if convergence time is an issue.

^[3] The definition of small, medium, and large IP networks can be discussed ad nauseam because of the number of variables involved (number of routers and routes, network bandwidth/utilization, network delay/latency, etc.), but rough measures are as follows: small -- a few ...