When routes will not summarize: discontiguous networks explained
Route summarization is a cornerstone of scalable network design. It reduces routing table size, lowers convergence time, and conserves bandwidth on WAN links. But summarization has hard limits, and the most common blocker is a discontiguous network topology. When your subnets are discontiguous, you cannot summarize them without losing reachability to parts of your network. Understanding why this happens is essential for network design and troubleshooting.
What is a discontiguous network?
A discontiguous network occurs when subnets belonging to the same major classful address space are separated by subnets from a different major address space. In other words, the same Class A, B, or C network appears in two or more separate locations in your routing topology, with different networks in between.
For example, imagine you have 10.1.0.0 divided by 24 (10.1.1.0/24, 10.1.2.0/24) connected to Router A, and 10.1.3.0/24 connected to Router B on the far side of a 172.16.0.0 divided by 16 WAN link. The 10.1.0.0 divided by 16 major network is now discontiguous because it is split across two routers with a different major network in between.
Why summarization breaks with discontiguous networks
Route summarization works by finding the smallest supernet that covers all your subnets. The summarized route is advertised instead of individual routes. But summarization assumes all subnets within that supernet are reachable via the same path.
With discontiguous networks, this assumption fails. If you summarize 10.1.0.0 divided by 22 to cover 10.1.0.0/24 and 10.1.1.0/24 and 10.1.2.0/24 and 10.1.3.0/24, the supernet also includes 10.1.2.0/24 and 10.1.3.0/24. If those subnets are actually behind a different router, traffic destined for them will be sent to the wrong next hop. Packets will be dropped or blackholed.
Classful routing protocols like RIPv1 and IGRP make this worse. They automatically summarize at major network boundaries and have no way to know about discontiguous topologies. Modern classless protocols like OSPF, IS-IS, BGP, and RIPv2 allow you to disable auto-summarization, but you must still avoid summarizing discontiguous subnets manually.
Common scenarios and how to identify them
- →Mergers and acquisitions: two sites both use 10.0.0.0 divided by 8, connected via a third site using 192.168.0.0 divided by 16
- →Legacy network expansion: older subnets (10.1.0.0/24) coexist with newer subnets (10.3.0.0/24) separated by intermediate routers
- →Multi-protocol environments: some segments use 10.0.0.0 divided by 8, others use 172.16.0.0 divided by 12, with 10.x.x.x appearing again downstream
- →Incomplete network redesign: partial migration from one address plan to another, leaving old and new ranges in the same major network
To identify discontiguous networks in your topology, draw a map of all subnets and their connections. Look for the same major classful network (Class A, B, or C) appearing in two or more disconnected regions. Use show ip route and show ip bgp commands to inspect your routing table. If you see overlapping or unexpected routes, or if traffic to certain subnets fails intermittently, discontiguous networks may be the cause.
Solutions and best practices
- →Redesign your address space: use different major networks for each region or site to eliminate overlap
- →Disable auto-summarization: in RIPv2, EIGRP, and BGP, turn off automatic summarization at major boundaries
- →Use classless routing protocols: OSPF, IS-IS, and BGP support CIDR and do not auto-summarize
- →Advertise specific routes: instead of summarizing, advertise individual subnets from discontiguous regions
- →Plan ahead: during network design, ensure each major network is contiguous or use entirely separate address spaces
The cleanest solution is to redesign your address space during a planned migration. Assign each geographic region or autonomous system its own major network block. This eliminates discontiguity, enables efficient summarization, and simplifies troubleshooting.