Routing protocols maintain routing tables. Each router has a routing table for each Layer 3 routed protocol it supports. To maintain current information about the conditions of the paths to which it connects, a routing protocol communicates with other routers within its vicinity. It then uses the information it learns to update its routing table. This is how routers make a path determination forwarding or filtering decision.

Routing Information Base (RIB) is the actual routing table that is built by a routing protocol, such as RIP, OSPF, and others. A RIB entry points to the destination network number and associated router interface. A forwarding information base (FIB) is the listing of the next-hop router for each destination prefix or network number. The FIB is constructed from the RIB and provides the specific interface details that the destination network is aligned towards.

An example of an international routing number is the E.164 international telecommunications number plan, which identifies a unique country code or routing number for each international country destination. This is similar to how a destination network number is identified in an IP routing table.

Routing Tables. A router is connected to at least two networks—its own network and at least one external network (such as an Internet service provider’s [ISP’s] gateway router). The router’s primary job is to determine which of its connections are best to use for a packet to reach its destination most efficiently. Because the network environment is constantly changing, this data must be updated frequently so it reflects the true and current condition of the network. In a network where a router has only a few entries in its routing table, forwarding decisions are relatively simple. In a large network, where multiple routers each have multiple links, the routing tables are much larger, and forwarding decisions can be more complex.

Each entry in a routing table specifies one particular route (path or link). All routing tables should contain at least one entry—a default route. A default route is the specified route that an IP packet should take from the source device when no other specific route is identified. An IP default gateway router is the router that acts as that workstation’s gateway to exit the LAN. This entry contains the information of the router’s default gateway, which is typically the gateway address of the WAN interface or next hop address of the service provider; however, a default gateway can also be the address where all packets are forwarded if no entry exists in the routing table for its destination.

A routing table entry, which can vary depending on the routed protocol in use, typically consists of the following information:

• Network ID or host internetwork logical address
• Subnet or prefix mask to extract the network ID from the destination IP address
• Forwarding address
• Interface port associated with the network ID
• Routing metric

If the destination address of the packet to be forwarded matches an entry in the routing table, the metric value is examined to determine whether this route has the lowest cost (meaning the lowest metric value of the available routes). If multiple routes are available, the route with the lowest cost is chosen, and the packet is forwarded on its link. If only one route is available, it is used.

Entries in a routing table, which is kept in the router’s main memory, reference directly connected networks and/or remotely connected networks. Directly connected networks are those in which one router is linked to another router by a direct connection. A message routed to a remotely connected router must pass through an intermediary router.

How a Routing Table Works. When a router receives a forwarded packet, it extracts the packet’s destination address and determines which interface port the packet should be forwarded to, based on the routing table. In the case where a router has only two links, the decision is an either/or. In cases where a router has multiple paths to the destination, the better route for the packet must be determined from the information in the routing table that contains a separate entry for each of the links.

Some Common Routing Protocols. A few of the more commonly used routing protocols are listed in TABLE 8-1. More information is available on each of the routing protocols later in this chapter.

Running a dual stack routing process means the router has both IPv4 and IPv6 enabled, providing a network migration path between a legacy IPv4 network domain and a newer IPv6 network domain. An intermediate router that has both IPv4 and IPv6 enabled is called a dual stack router.

A routed protocol defines the structure and format of the data being forwarded by a router and its routing protocol. Once a packet is formatted in accordance with the packet’s routed protocol, the packet can be forwarded by the routing protocol. Examples of routed protocols are IP, Novell Internetwork Packet Exchange (IPX), DECnet, AppleTalk, and Banyan Vines, as well as subprotocols of IP such as Telnet, Remote Procedure Call (RPC), Simple Network Management Protocol (SNMP), and Simple Mail Transfer Protocol (SMTP).

Some protocols cannot be routed. These protocols assume that all of the hosts with which they must communicate speak the same language and are located on the same network. For networks running nonroutable protocols to communicate, a network bridge or Layer 2 networking is required. Examples of nonroutable protocols are Dynamic Host Configuration Protocol (DHCP), NetBIOS Extended User Interface (NetBEUI), Data Link Command (DLC), Local Area Transport (LAT), Director Response Protocol (DRP), and Maintenance Operation Protocol (MOP). These protocols do not have a Network Layer address; hence, they are nonroutable protocols that must operate at Layer 2.

It is important to understand the difference between routing protocols and routed protocols. Routing protocols are used to propagate dynamically learned route information from other routers participating in the same process. Routed protocols are responsible for the delivery of the datagram and contain the Network Layer addressing. The routing protocols use this Network Layer information to determine the best path for the packet to take across the network to reach its destination. For example, IP is a routed protocol, and inside the IP packet resides its source and destination network (Layer 3) addresses. A routing protocol, such as the Routing Information Protocol (RIP), Open Shortest Path First (OSPF), or Border Gateway Protocol (BGP), covered later in this text in more detail, will inspect the destination network address inside the packet. It will make a routing decision on where to forward (route) the packet based on the information in its routing table.

Routing protocols talk to neighbor routers to learn about networks and the interfaces they are associated with. Neighbor activity refers to the dialogue that neighbor routers and their interconnected interfaces perform. Keep-alives, neighbor discovery, and the MAC layer address information of neighbor interfaces are identified via this neighbor activity.

Distributed routing or switching is when you have a central core router or switch, interconnected to outer-lying routers and switches. In this scenario, multiple processors are making the routing or switching decision based on a distributed architecture. Distributed switching typically involves some kind of hierarchy of routers or switches in the design, allowing for switching to be active and closer to the endpoints.

In an IP-based environment, every node, or device, will have its own individual IP address. Devices such as routers, which have an interface on each network involved, will have multiple IP addresses or an IP address per interface. A solid understanding of subnet masks and how they determine the network and host ID of the IP address is vital for learning about routing and path determination.

Before getting into more detailed examples of routing, here are the basics of how routing tables work and how path determination is made. In the example shown in FIGURE 8-2, Router A has an interface on Network 1, to which Switch A and Host A belong. Router B has an interface on Network 3, to which Switch B and Host B belong. Network 2 is the transit network, or the network that is used to pass packets between Networks 1 and 3. Each router has an interface on Network 2. In this example, if you assume Host A needs to talk to Host B, then Host A will forward its packet out of its default gateway because Host B does not live on the same network as Host A. The default gateway in this case is the interface of Router A on Network 1. It’s important to remember that in topologies such as these, the switch does not act as a Layer 3 device and will just pass packets on to hosts on the same network or to the router. Router A receives the packet and references its routing table. The destination network, Network 3, isn’t directly connected as are Networks 1 and 2. Instead, Router A sees Network 3 in its routing table as an external route, known via Router B. Router A knows, via its routing table, that to get to Network 3, it must pass the packet on to the next hop IP address. This is the interface of Router B on the transit network, Network 2. Router B will then check its routing table and determine that it’s a directly connected route and forward the packet on to Host B.

In this scenario, Router A originally knows only about its directly connected networks, Networks 1 and 2. Router B likewise knows about its directly connected networks, Networks 2 and 3. There are two ways for these routers to communicate with each other. One is to configure a dynamic routing protocol, such as one of the examples previously mentioned (RIP, OSPF, BGP, and so on). A dynamic routing protocol learns about its directly connected routes and then forwards that information to its directly connected routers, or in more complex environments, to other routers configured in the same process. In FIGURE 8-2, if you were using a distance vector protocol such as RIP, Router A would forward a copy of its routing table (Networks 1 and 2) to Router B. Router B would do the same, passing a copy of its routing table (Networks 2 and 3) to Router A. Router A would see that Network 3 is from an external router and install it in its routing table. Any future packet forwarded to Router A, destined to Network 3, it would know to forward to Router B.

There is another way to route packets and manipulate path determination without the use of dynamic routing protocols. This method is called static routing. Static routing is typically only used in smaller environments with a network that doesn’t experience much change. As opposed to dynamic routing protocols that share information with other routers, static routing remains locally significant to the router. Static routing is performed on the router by a network administrator manually configuring static routes. These routes tell the router which interface or next hop router to use to get to a specific destination network. If a router loses a link to another router, these static routes must be changed manually because there is no automatic network convergence feature such as is found in dynamic routing protocols.

To show how static routing works, refer to FIGURE 8-2. In this example, if Router A needs to get packets to Network 3, then a static route must be installed in Router A that says, “To get to Network 3, forward packets to Router B.” If you set up this environment in a lab and configured that previous static route and had sniffers, or network capture devices, on each side, you would see the traffic from Router A reaching Host B. The problem is that the return traffic would fail, because Router B would not know how to pass the return traffic. Therefore, a static route also must be configured on Router B that says “To get to Network 1, forward packets to Router A.” This would allow full communication between the two hosts.

Now that you have learned how routing tables are updated, let’s look at an example with multiple paths. FIGURE 8-3 shows a three-router network with Router A connected to networks 192.168.10.0, 192.168.20.0, and 192.168.30.0; Router B is connected to networks 192.168.30.0, 192.168.40.0, 192.168.50.0, and 192.168.60.0; and Router C is connected to networks 192.168.60.0, 192.168.10.0, and 192.168.70.0. Notice that to each router, the networks to which it’s attached are recognized by the IP address assigned to the interface port to which the network is attached. Each interface recognizes and supports the protocols of each link and can therefore determine the state of each network (up or down).

To build its routing table:

1. Router A uses the subnet masks associated with each interface to determine that it’s attached to networks 192.168.10.0, 192.168.20.0, and 192.168.30.0.
2. Router A then adds an entry for each network into its routing table along with the information that each network is directly connected.
3. Router A then sends routing update packets to Routers B and C informing them that it’s directly connected to these three networks.
4. Routers B and C perform the same steps and send routing updates to Router A informing it of their directly connected networks.
5. Router A updates its routing table with the source address included in each router’s routing update packet and the network addresses provided by each router.

However simple this may sound, it can become more complicated and troublesome:

• Both Routers B and C reported being directly connected to 192.168.60.0. Which is the better choice for reaching that network, or are they equals?
• Should Router A update Router C about the networks reported from Router B? What if the connection between them is down? Should Router A pass that information along?

These and other questions concerning the status, availability, and path to various networks require additional information before they can be answered. This is where routing metrics come in.

If you assume each router has shared routing information with the other two routers in FIGURE 8-3, you know that certain networks appear to have multiple paths to them. A routing metric is a calculation of the cost for a given path. That cost can vary depending on the routing protocol in use and parameters such as lowest hop count, lowest latency, highest bandwidth, lowest load, and path reliability. A metric is a value calculated for each route that ranks the available routing options. The router then chooses, based on the best metric value, which path to take.

TABLE 8-2 shows the beginnings of an entry in Router A’s routing table. This entry shows that to reach network 192.168.50.0, Router A has two distinct routing options: the directly connected links to each of Router B and Router C.

At this point, the routing table entry contains no information indicating which of the two routing options is the better path to use. A metric stored for each of the routing options that provides a relative ranking is needed. The criteria and meaning of a routing metric vary with each of the different routing protocols. Routing protocols use two main parameters: hop count and bandwidth.

Not all routing protocols use the same metrics. The Routing Information Protocol (RIP) bases its “best” route decisions strictly on the lowest hop count. Open Shortest Path First (OSPF) uses a metric that is reference bandwidth (100 Mbps) divided by the interface bandwidth (which is value 1 for 100 Mbps through 100 Gbps; slower interface speeds have larger values).

Hop Count. A hop count metric indicates the number of intermediary routers a packet will have to pass through to reach a destination network. The hop count is just a count of the number of times a packet must be forwarded by other routers to reach its destination. Each time a packet is forwarded, it counts as a hop.

Looking back at the network in FIGURE 8-3 and the information in Table 8-2, if hop count is added to the information, the hop count for a packet addressed to a host on the 192.168.50.0 network and passing through Router C is two. One hop is to Router C’s 192.168.10.2 interface, and the second hop is to Router B through the 192.168.60.1 interface. The hop count for this packet through Router B is one, a direct connected link, 192.168.30.1.

Therefore, based strictly on the hop count metric, Router A would update its routing table with this information, as shown in TABLE 8-3. As long as this information remains true, and hop count is the only metric in use, Router A will forward packets for the 192.168.50.0 network to Router B.

Bandwidth. Sometimes hop count doesn’t really tell the whole story of a link. Perhaps the link from Router A to Router B (see FIGURE 8-3) is a dial-up line, and the link between Router A and Router C is a dedicated leased T-1 line. In this exaggerated example, it’s relatively easy to say that the two-hop link actually may be the better path. Bandwidth and the condition of the link are other ways to determine the quality of a routing option. A routing protocol that uses a bandwidth metric chooses a path with higher bandwidth over one with lower bandwidth, regardless of the type of link. Bandwidth, all by itself, may not be a single indicator of the condition and quality of a link. It could be that a 56 kbps link is very lightly loaded, and the T-1 link is seeing heavy use at any given moment.

In addition to the bandwidth of a particular link, other metrics can factor in to determine the overall desirability of the link, especially in comparison to other available links. The other metrics that may be included in this determination, when using EIGRP, are:

• Load—On any particular route, hop count and bandwidth generally stay the same. However, the amount of traffic or load, which is the amount of bandwidth in use on the link, can fluctuate quickly. A routing protocol that depends too heavily on the line loading for its routing determination may end up with what is called route flapping, rapidly changing the preferred route from one route to another and perhaps back again. Route flapping can be caused by hardware, software or configuration errors, or unreliable connections, which cause availability information to be repeatedly advertised and withdrawn. However, the idea behind using a load metric is that the best path is the one with the lightest load.

• Delay—A delay metric essentially measures a link’s throughput or transit time: the duration time for a packet to completely navigate a link end-to-end. A routing protocol using a delay metric favors the link with the least delay. A delay metric can be developed a variety of ways: clocking the time between when a packet is forwarded until the time the packet is acknowledged, for example. Alternatively, it could be the sum of the delay expected from the types of media along the link and any delay caused by latency of the routers and queuing that occurs along the route. Or delay can be just a time estimate based solely on the link’s medium.
• Reliability—The reliability metric is an indicator of how likely it is that the link may fail during a transmission. Reliability metrics are either variable or fixed. A variable reliability metric may be a count of the actual number of failures that have occurred on a link within a specific time. A fixed reliability metric is a static metric that is typically manually configured by the network administrator and is based on his or her knowledge of a particular link. Of course, the path with the highest indication of reliability would be the preferred link.
• Maximum transmission unit (MTU)—MTU has been called the fifth factor in deciding the EIGRP metric. A change in the MTU doesn’t trigger a routing update. When a change in topology triggers an EIGRP update, the MTU metric updates. If the router determines there is a tie for a routing decision, the MTU is used as the tie-breaker.

Cost. In the context of routing, a cost metric doesn’t refer to dollars and cents, but rather to the relative metric value that is used to set the condition of a comparison. For example, the route with the lowest cost is likely the link that has the lowest overall value for the metrics included in the path determination method of a protocol. Any path chosen as a preferred route will be the routing option with the lowest cost.