Route Preferences Overview

Autonomous Systems

A large network or collection of routers under a single administrative authority is termed an autonomous system (AS). Autonomous systems are identified by a unique numeric identifier that is assigned by the Internet Assigned Numbers Authority (IANA). Typically, the hosts within an AS are treated as internal peers, and hosts in a peer AS are treated as external peers. The status of the relationship between hosts—internal or external—governs the protocol used to exchange routing information.

Alternate and Tiebreaker Preferences

The Junos OS provides support for alternate and tiebreaker preferences, and some of the routing protocols, including BGP and label switching, use these additional preferences. With these protocols, you can specify a primary route preference (by including the preference statement in the configuration), and a secondary preference that is used as a tiebreaker (by including the preference2 statement).

In order to use common comparison routines, Junos OS stores the 1's complement of the LocalPref value in the Preference2 field. For example, if the LocalPref value for Route 1 is 100, the Preference2 value is -101. If the LocalPref value for Route 2 is 155, the Preference2 value is -156. Route 2 is preferred because it has a higher LocalPref value and a lower Preference2 value.

You can also mark route preferences with additional route tiebreaker information by specifying a color and a tiebreaker color (by including the color and the tiebreaker color2 statements in the configuration). color and color2 statements are even finer-grained preference values that Junos OS uses when preference and preference2 statements fail to break the tie during route selection.

The software uses a 4-byte value to represent the route preference value. When using the preference value to select an active route, the software first compares the primary route preference values, choosing the route with the lowest value. If there is a tie and a secondary preference has been configured, the software compares the secondary preference values, choosing the route with the lowest value. The secondary preference values must be included in a set for the preference values to be considered.

Multiple Active Routes

The IGPs compute equal-cost multipath next hops, and IBGP picks up these next hops. When there are multiple, equal-cost next hops associated with a route, the routing protocol process installs only one of the next hops in the forwarding path with each route, randomly selecting which next hop to install. For example, if there are 3 equal-cost paths to an exit routing device and 900 routes leaving through that routing device, each path ends up with about 300 routes pointing at it. This mechanism provides load distribution among the paths while maintaining packet ordering per destination.

BGP multipath does not apply to paths that share the same MED-plus-IGP cost yet differ in IGP cost. Multipath path selection is based on the IGP cost metric, even if two paths have the same MED-plus-IGP cost.

Random selection of equal-cost multipath occurs independently for inet.0 and inet.3 tables. This can lead to a single prefix showing different bestpaths for inet.0 vs inet.3.

Dynamic and Static Routing

Entries are imported into a router's routing table from dynamic routing protocols or by manual inclusion as static routes. Dynamic routing protocols allow routers to learn the network topology from the network. The routers within the network send out routing information in the form of route advertisements. These advertisements establish and communicate active destinations, which are then shared with other routers in the network.

Although dynamic routing protocols are extremely useful, they have associated costs. Because they use the network to advertise routes, dynamic routing protocols consume bandwidth. Additionally, because they rely on the transmission and receipt of route advertisements to build a routing table, dynamic routing protocols create a delay (latency) between the time a router is powered on and the time during which routes are imported into the routing table. Some routes are therefore effectively unavailable until the routing table is completely updated, when the router first comes online or when routes change within the network (due to a host going offline, for example).

Static routing avoids the bandwidth cost and route import latency of dynamic routing. Static routes are manually included in the routing table, and never change unless you explicitly update them. Static routes are automatically imported into the routing table when a router first comes online. Additionally, all traffic destined for a static address is routed through the same router. This feature is particularly useful for networks with customers whose traffic must always flow through the same routers. Figure 1 shows a network that uses static routes.

Figure 1: Static Routing Example

In Figure 1, the customer routes in the 192.176.14/24 subnetwork are static routes. These are hard links to specific customer hosts that never change. Because all traffic destined for any of these routes is forwarded through Router A, these routes are included as static routes in Router A's routing table. Router A then advertises these routes to other hosts so that traffic can be routed to and from them.

Route Advertisements

The routing table and forwarding table contain the routes for the routers within a network. These routes are learned through the exchange of route advertisements. Route advertisements are exchanged according to the particular protocol being employed within the network.

Generally, a router transmits hello packets out each of its interfaces. Neighboring routers detect these packets and establish adjacencies with the router. The adjacencies are then shared with other neighboring routers, which allows the routers to build up the entire network topology in a topology database, as shown in Figure 2.

Figure 2: Route Advertisement

In Figure 2, Router A sends out hello packets to each of its neighbors. Routers B and C detect these packets and establish an adjacent relationship with Router A. Router B and C then share this information with their neighbors, Routers D and E, respectively. By sharing information throughout the network, the routers create a network topology, which they use to determine the paths to all possible destinations within the network. The routes are then distilled into the forwarding table of best routes according to the route selection criteria of the protocol in use.

Route Aggregation

As the number of hosts in a network increases, the routing and forwarding tables must establish and maintain more routes. As these tables become larger, the time routers require to look up particular routes so that packets can be forwarded becomes prohibitive. The solution to the problem of growing routing tables is to group (aggregate) the routers by subnetwork, as shown in Figure 3.

Figure 3: Route Aggregation

Figure 3 shows three different ASs. Each AS contains multiple subnetworks with thousands of host addresses. To allow traffic to be sent from any host to any host, the routing tables for each host must include a route for each destination. For the routing tables to include every combination of hosts, the flooding of route advertisements for each possible route becomes prohibitive. In a network of hosts numbering in the thousands or even millions, simple route advertisement is not only impractical but impossible.

By employing route aggregation, instead of advertising a route for each host in AS 3, the gateway router advertises only a single route that includes all the routes to all the hosts within the AS. For example, instead of advertising the particular route 170.16.124.17, the AS 3 gateway router advertises only 170.16/16. This single route advertisement encompasses all the hosts within the 170.16/16 subnetwork, which reduces the number of routes in the routing table from 2¹⁶ (one for every possible IP address within the subnetwork) to 1. Any traffic destined for a host within the AS is forwarded to the gateway router, which is then responsible for forwarding the packet to the appropriate host.

Similarly, in this example, the gateway router is responsible for maintaining 2¹⁶ routes within the AS (in addition to any external routes). The division of this AS into subnetworks allows for further route aggregation to reduce this number. In the subnetwork in the example, the subnetwork gateway router advertises only a single route (170.16.124/24), which reduces the number of routes from 2⁸ to 1.