This section covers PBR, route summarization, and route redistribution. You can use PBR to modify the next hop of packets from what is selected by the routing protocol. PBR is useful when the traffic engineering of paths is required. Routes are summarized to reduce the size of routing tables and at network boundaries. Redistribution between routing protocols is required to inject route information from one routing protocol to another. The CCDA must understand the issues with the redistribution of routes.
You can use PBR to modify the next-hop address of packets or to mark packets to receive differential service. Routing is based on destination addresses; routers look at the routing table to determine the next-hop IP address based on a destination lookup. PBR is commonly used to modify the next-hop IP address based on the source address. You can also use PBR to mark the IP precedence bits in outbound IP packets so that you can apply quality-of-service (QoS) policies. In Figure 12-9, Router A exchanges routing updates with routers in the WAN. The routing protocol might select Serial 0 as the preferred path for all traffic because of the higher bandwidth. The company might have business-critical systems that use the T1 but does not want systems on Ethernet 1 to affect WAN performance. You can configure PBR on Router A to force traffic from Ethernet 1 out on Serial 1.
Figure 12-9 Policy-Based Routing
Large networks can grow very quickly from 500 routes to 1000, to 2000, and so on. Network IP addresses should be allocated to allow for route summarization. Route summarization reduces the amount of route traffic on the network and unnecessary route computation. Route summarization also allows the network to scale as a company grows.
The recommended location for route summarization is to summarize at the distribution layer of the network topology. Figure 12-10 shows a hierarchical network. It has a network core, regional distribution routers, and access routes for sites.
Figure 12-10 Route Summarization
All routes in Brazil are summarized with a single 10.1.0.0/16 route. The North America and European routes are also summarized with 10.2.0.0/16 and 10.3.0.0/16, respectively. Routers in Europe only need to know the summarized route to get to Brazil and North America, and vice versa. Again, design best practices are to summarize at the distribution toward the core. The core only needs to know the summarized route of the regional areas.
You configure the redistribution of routing protocols on routers that reside at the Service Provider Edge of the network. These routers exchange routes with other autonomous systems. Redistribution is also done on routers that run more than one routing protocol. Here are some reasons to do redistribution:
- Migration from an older routing protocol to a new routing protocol.
- Mixed-vendor environment in which Cisco routers might be using EIGRP and other vendor routers might be using OSPF.
- Different administrative domain between company departments using different routing protocols.
- Mergers and acquisitions in which the networks initially need to communicate. In this example two different EIGRP processes might exist.
Figure 12-11 shows an example of the exchange of routes between two autonomous systems. Routes from AS 100 are redistributed into BGP on Router A. Routes from AS 200 are redistributed into BGP on Router B. Then, Routers A and B exchange BGP routes. Router A and Router B also implement filters to redistribute only the desired networks.
Figure 12-11 Redistribution of BGP Routes
A company might also acquire another company that might be running another routing protocol. Figure 12-12 shows a network that has both OSPF and EIGRP routing protocols. Routers A and B perform redistribution between OSPF and EIGRP. Both routers must filter routes from OSPF before redistributing them into EIGRP and filter routes from EIGRP before redistributing them into OSPF. This setup prevents route feedback.
Figure 12-12 Redistribution Between IGPs
Route feedback occurs when a routing protocol learns routes from another routing protocol and then announces the routes to the other routing protocol. In Figure 12-12, OSPF should not announce the routes it learned from EIGRP, and EIGRP should not announce the routes it learned from OSPF.
You can use access lists, distribution lists, and route maps when redistributing routes. You can use these methods to specify (select) routes for redistribution, to set metrics, or to set other policies for the routes. They are also used to control routes' redistribution direction. Redistribution can be accomplished by two methods:
- Two-way redistribution
- One-way redistribution
In two-way redistribution, routing information is exchanged between both routing protocols. No static routes are used in this exchange. Route filters are used to prevent routing loops. Routing loops can be caused by one route protocol redistributing routes that were learned from a second route protocol back to that second routing protocol.
One-way redistribution only allows redistribution from one routing protocol to another. Normally it is used in conjunction with a default or static route at the edge of a network. Figure 12-13 shows an example of one-way redistribution. The routing information from the WAN routes is redistributed into the campus. But campus routes are not redistributed out to the WAN. The WAN routers use a default gateway to get back to the campus.
Figure 12-13 One-Way Route Redistribution
Other locations for one-way redistribution are from building access networks, BGP routes or static routes into the IGP, and from VPN static routes into the IGP.
There is a default metric of 0 when redistributing routes into RIPv2, IS-IS, and EIGRP. You should configure the metric of the redistributed routes to a metric other than 0. You can configure the metric in the redistribution command or configure a default metric. You can also use the command in OSPF. IS-IS does not use the default-metric command. The default-metric command has the following syntax for EIGRP:
default-metric bandwidth delay reliability load mtu
This subsection reviews a few things you need to remember when designing a network that will redistribute with OSPF.
When redistributing routes into OSPF, use the subnets keyword to permit subnetted routes to be received. If you do not use it, only the major network route is redistributed, without any subnetworks. In other words, OSPF performs automatic summarization to IP classful network values.
By default, redistributed routes are classified as external Type 2 (E2) in OSPF. You can use the metric-type keyword to change the external route to an external Type 1 (E1). The network design can take into account the after-redistribution cost (Type 2) or the after-redistribution cost plus the path's cost (Type 1).
In Figure 12-14, Router B is configured to perform mutual redistribution between EIGRP 100 and OSPF process ID 50. In this example, you can use route maps and access lists to prevent routing loops. The route maps permit or deny the networks that are listed in the access lists. The subnets keyword redistributes every subnet in EIGRP into OSPF. This book does not cover exact configurations.
Figure 12-14 OSPF and EIGRP Redistribution