Static Versus Dynamic Route Assignment

Static routes are manually configured on a router. When configured manually and not learned from a neighbor, static routes do not react to network outages. The one exception is when a static route specifies the outbound interface or the IP address of the next hop is not resolved in the routing table. In this situation, if the interface goes down or the next hop IP address is not resolved in the routing table, the static route is removed from the routing table. Because static routes are unidirectional, they must be configured for each outgoing interface the router will use. The size of today’s networks makes it impossible to manually configure and maintain all the routes in all the routers in a timely manner. Human configuration can involve many mistakes. Dynamic routing protocols were created to address these shortcomings. They use algorithms to advertise, learn about, and react to changes in the network topology.

The main benefit of static routing is that a router generates no routing protocol overhead. Because no routing protocol is enabled, no bandwidth is consumed by route advertisements between network devices. Another benefit of static routing is that static routing protocols are easier to configure and troubleshoot than dynamic routing protocols. Static routing is recommended for hub-and-spoke topologies with low-speed remote connections and where only a single path to the network exists. A default static route is configured at each remote site because the hub is the only route used to reach all other sites. Static routes are also used at network boundaries (the Internet or partners) where routing information is not exchanged. These static routes are then redistributed into the internal dynamic routing protocol used.

Figure 3-1 shows a hub-and-spoke WAN with static routes defined in the remote WAN routers because no routing protocols are configured. This setup eliminates routing protocol traffic on the low-bandwidth WAN circuits.

Routing protocols dynamically determine the best route to a destination. When the network topology changes, the routing protocol adjusts the routes without administrative intervention. Routing protocols use a metric to determine the best path toward a destination network. Some use a single measured value such as hop count. Others compute a metric value using one or more parameters. Routing metrics are discussed later in this chapter. The following is a list of dynamic routing protocols:

• RIPv1

• RIPv2

• EIGRP

• OSPF

• IS-IS

• RIPng

• OSPFv3

• EIGRP for IPv6

• Border Gateway Protocol (BGP)

Interior Versus Exterior Routing Protocols

Routing protocols can be categorized as interior gateway protocols (IGPs) or exterior gateway protocols (EGPs). IGPs are meant for routing within an organization’s administrative domain (in other words, the organization’s internal network). EGPs are routing protocols used to communicate with exterior domains, where routing information is exchanged between administrative domains. Figure 3-2 shows where an internetwork uses IGPs and EGPs with multiple autonomous administrative domains. BGP exchanges routing information between the internal network and an ISP. IGPs appear in the internal private network.

One of the first EGPs was called exactly that: Exterior Gateway Protocol. Today, BGP is the de facto (and the only available) EGP.

Potential IGPs for an IPv4 network are

• RIPv2

• OSPFv2

• IS-IS

• EIGRP

Potential IGPs for an IPv6 network are

• RIPng

• OSPFv3

• EIGRP for IPv6

• IS-IS

RIPv1 is no longer recommended because of its limitations. RIPv2 addresses many of the limitations of RIPv1 and is the most recent version of RIP. IGRP is an earlier version of EIGRP. RIPv1, RIPv2, and IGRP are not included on the ENSLD 300-420 exam blueprint. Table 3-2 provides a quick high-level summary of routing protocol uses.

Distance-Vector Routing Protocols

The first IGP routing protocols introduced were distance-vector routing protocols. They used the Bellman-Ford algorithm to build routing tables. With distance-vector routing protocols, routes are advertised as vectors of distance and direction. The distance metric is usually router hop count. The direction is the next-hop router (IP address) toward which to forward the packet. For RIP, the maximum number of hops is 15, which can be a serious limitation, especially in large nonhierarchical internetworks.

Distance-vector algorithms call for each router to send its entire routing table to only its immediate neighbors. The table is sent periodically (every 30 seconds for RIP). In the period between advertisements, each router builds a new table to send to its neighbors at the end of the period. Because each router relies on its neighbors for route information, it is commonly said that distance-vector protocols “route by rumor.” A 30 second wait for a new routing table with new routes is too long for today’s networks. This is why distance-vector routing protocols have slow convergence.

RIPv2 and RIPng can send triggered updates—that is, full routing table updates sent before the update timer has expired. A router can receive a routing table with 500 routes with only one route change, which creates serious overhead on the network (another drawback). Furthermore, RFC 2091 updates RIP with triggered extensions to allow triggered updates with only route changes. Cisco routers support this on fixed point-to-point interfaces.

The following is a list of IP distance-vector routing protocols:

• RIPv1 and RIPv2

• EIGRP (which could be considered a hybrid)

• RIPng

Link-State Routing Protocols

Link-state routing protocols address some of the limitations of distance-vector protocols. When running a link-state routing protocol, routers originate information about themselves (IP addresses), their connected links (the number and types of links), and the state of those links (up or down). The information is flooded to all routers in the network as changes in the link state occur. Each router makes a copy of the information received and forwards it without change. Each router independently calculates the best path to each destination network by using Dijkstra’s shortest path algorithm, creating a shortest path tree with itself as the root, and maintaining a map of the network.

After the initial exchange of information, link-state updates are not sent unless a change in the topology occurs. Routers do send small hello messages between neighbors to maintain neighbor relationships. If no updates have been sent, the link-state route database is refreshed after 30 minutes.

The following is a list of link-state routing protocols:

• OSPFv2

• IS-IS

• OSPFv3

OSPFv2 and OSPFv3 are covered in Chapter 4, “OSPF, BGP, and Route Manipulation.”

Distance-Vector Routing Protocols Versus Link-State Protocols

When choosing a routing protocol, consider that distance-vector routing protocols use more network bandwidth than link-state protocols. Distance-vector protocols generate more bandwidth overhead because of the large periodic routing updates. Link-state routing protocols do not generate significant routing update overhead but do use more router CPU and memory resources than distance-vector protocols. This occurs because with link-state routing protocols (generally speaking), WAN bandwidth is a more expensive resource than router CPU and memory in modern devices.

Table 3-3 compares distance-vector and link-state routing protocols.

EIGRP is a distance-vector protocol with link-state characteristics (hybrid) that give it high scalability, fast convergence, less routing overhead, and relatively easy configuration. If “distance-vector” is not an answer to a question on the ENSLD 300-420 exam, then “hybrid” would be a valid option.

Hierarchical Versus Flat Routing Protocols

Some routing protocols require a network topology that must have a backbone network defined. This network contains some or all of the routers in the internetwork. When the internetwork is defined hierarchically, the backbone consists of only some devices. Backbone routers service and coordinate the routes and traffic to or from routers not in the local internetwork. The supported hierarchy is relatively shallow. Two levels of hierarchy are generally sufficient to provide scalability. Selected routers forward routes into the backbone. OSPF and IS-IS are hierarchical routing protocols. By default, EIGRP is a flat routing protocol, but it can be configured with manual summarization to support hierarchical designs.

Flat routing protocols do not allow a hierarchical network organization. They propagate all routing information throughout the network without dividing or summarizing large networks into smaller areas. Carefully designing network addressing to naturally support aggregation within routing-protocol advertisements can provide many of the benefits offered by hierarchical routing protocols. Every router is a peer of every other router in flat routing protocols; no router has a special role in the internetwork. EIGRP, RIPv1, and RIPv2 are flat routing protocols.

Classless Versus Classful Routing Protocols

Routing protocols can be classified based on their support of VLSM and CIDR. Classful routing protocols do not advertise subnet masks in their routing updates; therefore, the configured subnet mask for the IP network must be the same throughout the entire internetwork. Furthermore, the subnets must, for all practical purposes, be contiguous within the larger internetwork. For example, if you use a classful routing protocol for network 200.170.0.0, you must use the chosen mask (such as 255.255.255.0) on all router interfaces using the 200.170.0.0 network. You must configure serial links with only two hosts and LANs with tens or hundreds of devices with the same mask, 255.255.255.0. The big disadvantage of classful routing protocols is that the network designer cannot take advantage of address summarization across networks (CIDR) or allocation of smaller or larger subnets within an IP network (VLSM). For example, with a classful routing protocol that uses a default mask of /25 for the entire network, you cannot assign a /30 subnet to a serial point-to-point circuit. The following protocols are classful routing protocols:

• RIPv1

• IGRP (this protocol is not covered on the ENSLD 300-420 exam)

Classless routing protocols advertise the subnet mask with each route. You can configure subnetworks of a given IP network number with different subnet masks (using VLSM). You can configure large LANs with a smaller subnet mask and configure serial links with a larger subnet mask, thereby conserving IP address space. Classless routing protocols also allow flexible route summarization and supernetting (using CIDR). You can create supernets by aggregating classful IP networks. For example, 200.100.100.0/23 is a supernet of 200.100.100.0/24 and 200.100.101.0/24. The following protocols are classless routing protocols:

• RIPv2

• OSPF

• EIGRP

• IS-IS

• RIPng

• OSPFv3

• EIGRP for IPv6

• BGP

IPv4 Versus IPv6 Routing Protocols

With the increasing use of the IPv6 protocol, a CCNP enterprise designer must be prepared to design networks using IPv6 routing protocols. As IPv6 was defined, routing protocols needed to be updated to support the new IP address structure. None of the IPv4 routing protocols support IPv6 networks, and none of the IPv6 routing protocols are backward compatible with IPv4 networks. But both protocols can coexist on the same network, each with its own routing protocol. Devices with dual stacks recognize which protocol is being used by the IP Version field in the IP header.

RIPng is the IPv6-compatible RIP routing protocol. EIGRP for IPv6 is the new version of EIGRP that supports IPv6 networks. OSPFv3 was developed for IPv6 networks, and OSPFv2 remains for IPv4 networks. Internet drafts were written to provide IPv6 routing using IS-IS. Multiprotocol extensions for BGP provide IPv6 support for BGP. Table 3-4 lists IPv4 and IPv6 routing protocols.

Administrative Distance

On Cisco routers running more than one routing protocol, it is possible for two different routing protocols to have a route to the same destination. Cisco routers assign each routing protocol an administrative distance. When multiple routes exist for a destination, the router selects the longest match. For example, to reach the destination 170.20.10.1, if OSPF has a route prefix of 170.20.10.0/24, and EIGRP has a route prefix of 170.20.0.0/16, the OSPF route is preferred because the /24 prefix is longer than the /16 prefix. This means it is more specific.

If two or more routing protocols offer the same route (with the same prefix length) for inclusion in the routing table, the Cisco IOS router selects the route with the lowest administrative distance.

The administrative distance is a rating of the trustworthiness of a routing information source. Table 3-5 shows the default administrative distances for configured (static) or learned routes. In the table, you can see that static routes are trusted over dynamically learned routes. With IGP routing protocols, EIGRP internal routes are trusted over OSPF, IS-IS, and RIP routes.

The administrative distance establishes the precedence used among routing algorithms. Suppose a router has an internal EIGRP route to network 172.20.10.0/24 with the best path out Ethernet 0 and an OSPF route for the same network out Ethernet 1. Because EIGRP has an administrative distance of 90 and OSPF has an administrative distance of 110, the router enters the EIGRP route in the routing table and sends packets with destinations of 172.20.10.0/24 out Ethernet 0.

Static routes have a default administrative distance of 1. You can configure static routes with a different distance by appending the distance value to the end of the command.

Table 3-6 provides a summary of routing protocol characteristics.

Hop Count

The hop count parameter counts the number of links between routers that a packet must traverse to reach a destination. The RIP routing protocol uses hop count as the metric for route selection. If all links were the same bandwidth, this metric would work well. The problem with routing protocols that use only this metric is that the shortest hop count is not always the most appropriate path. For example, between two paths to a destination network—one with two 56 kbps links and another with four T1 links—the router chooses the first path because of the lower number of hops (see Figure 3-3). However, this is not necessarily the best path. You would prefer to transfer a 20 MB file via the T1 links rather than the 56 kbps links.

Bandwidth

The bandwidth parameter uses the interface bandwidth to determine the best path to a destination network. When bandwidth is the metric, the router prefers the path with the highest bandwidth to a destination. For example, a Fast Ethernet (100 Mbps) link is preferred over a DS-3 (45 Mbps) link. As shown in Figure 3-3, a router using bandwidth to determine a path would select Path 2 because of the greater bandwidth (1.5 Mbps over 56 kbps).

If a routing protocol uses only bandwidth as the metric and the path has several different speeds, the protocol can use the lowest speed in the path to determine the bandwidth for the path. EIGRP and IGRP use the minimum path bandwidth, inverted and scaled, as one part of the metric calculation. In Figure 3-4, Path 1 has two segments, with 256 kbps and 512 kbps of bandwidth. Because the smaller speed is 256 kbps, this speed is used as Path 1’s bandwidth. The smallest bandwidth in Path 2 is 384 kbps. When the router has to choose between Path 1 and Path 2, it selects Path 2 because 384 kbps is larger than 256 kbps.

Cost

Cost is the metric used by OSPF and IS-IS. In OSPF on a Cisco router, a link’s default cost is derived from the interface’s bandwidth. Cisco’s implementation of IS-IS assigns a default cost of 10 to all interfaces.

The formula to calculate cost in OSPF is

10⁸/BW

where BW is the interface’s default or configured bandwidth.

For 10 Mbps Ethernet, cost is calculated as follows:

BW = 10 Mbps = 10 × 10⁶ = 10,000,000 = 10⁷

cost (Ethernet) = 10⁸/10⁷ = 10

The sum of all the costs to reach a destination is the metric for that route. The lowest cost is the preferred path.

The path cost is the sum of all costs in the path. Figure 3-5 shows an example of how the path costs are calculated. The cost for Path 1 is 350 + 180 = 530. The cost for Path 2 is 15 + 50 + 100 + 50 = 215.

Because the cost of Path 2 is less than that of Path 1, Path 2 is selected as the best route to the destination.

Load

The load parameter refers to the degree to which the interface link is busy. A router keeps track of interface utilization; routing protocols can use this metric when calculating the best route. Load is one of the five parameters included in the definition of the EIGRP metric. By default, it is not used to calculate the composite metric. If you have 512 kbps and 256 kbps links to reach a destination, but the 512 kbps circuit is 99% busy and the 256 kbps circuit is only 5% busy, the 256 kbps link is the preferred path. On Cisco routers, the percentage of load is shown as a fraction over 255. Utilization at 100% is shown as 255/255, and utilization at 0% is shown as 0/255. Example 3-1 shows the load of a serial interface at 5/255 (1.9%).

Example 3-1 Interface Load

Click here to view code image

router3>show interface serial 1
Serial1 is up, line protocol is up
  Hardware is PQUICC Serial
  Internet address is 10.100.1.1/24
  MTU 1500 bytes, BW 1544 Kbit, DLY 20000 usec, rely 255/255, load 5/255

Delay

The delay parameter refers to how long it takes to move a packet to the destination. Delay depends on many factors, such as link bandwidth, utilization, port queues, and physical distance traveled. Total delay is one of the five parameters included in the definition of the EIGRP composite metric. By default, it is used to calculate the composite metric. You can configure an interface’s delay with the delay tens-of-microseconds command, where tens-of-microseconds specifies the delay, in tens of microseconds, for an interface or network segment. The interface delay can be checked with the show interface command. In Example 3-2, the interface’s delay is 20,000 microseconds.

Example 3-2 Interface Delay

Click here to view code image

router3>show interface serial 1
Serial1 is up, line protocol is up
  Hardware is PQUICC Serial
  Internet address is 10.100.1.1/24
  MTU 1500 bytes, BW 1544 Kbit, DLY 20000 usec, rely 255/255, load 1/255

Reliability

The reliability parameter is the dependability of a network link. Some WAN links tend to go up and down throughout the day. These links get a small reliability rating. Reliability is measured by factors such as a link’s expected received keepalives and the number of packet drops and interface resets. If the ratio is high, the line is reliable. The best rating is 255/255, which is 100% reliability. Reliability is one of the five parameters included in the definition of the EIGRP metric. By default, it is not used to calculate the composite metric. As shown in Example 3-3, you can verify an interface’s reliability by using the show interface command.

Example 3-3 Interface Reliability

Click here to view code image

router4#show interface serial 0
Serial0 is up, line protocol is up
  Hardware is PQUICC Serial
  MTU 1500 bytes, BW 1544 Kbit, DLY 20000 usec, rely 255/255, load 1/255

Maximum Transmission Unit

The MTU parameter is simply the maximum size of bytes a unit can have on an interface. If the outgoing packet is larger than the MTU, the IP protocol might need to fragment it. If a packet larger than the MTU has the Do Not Fragment flag set, the packet is dropped. As shown in Example 3-4, you can verify an interface’s MTU by using the show interface command.

Example 3-4 Interface MTU

Click here to view code image

router4#show interface serial 0
Serial0 is up, line protocol is up
  Hardware is PQUICC Serial
  MTU 1500 bytes, BW 1544 Kbit, DLY 20000 usec, rely 255/255, load 1/255

Routing Loop-Prevention Schemes

Some routing protocols employ schemes to prevent the creation of routing loops in the network. The following are the commonly used loop-prevention schemes:

• Split horizon

• Poison reverse

• Counting to infinity

Split Horizon

Split horizon is a technique used by distance-vector routing protocols to prevent routing loops. Routes that are learned from a neighboring router are not sent back to that neighboring router, thus suppressing the route. If the neighbor is already closer to the destination, it already has a better path.

In Figure 3-6, Routers 1, 2, and 3 learn about Networks A, B, C, and D. Router 2 learns about Network A from Router 1 and also has Networks B and C in its routing table. Router 3 advertises Network D to Router 2. Now, Router 2 knows about all networks. Router 2 sends its routing table to Router 3 without the route for Network D because it learned that route from Router 3.

Poison Reverse

Poison reverse is a route update sent out an interface with an infinite metric for routes learned (received) from the same interface. Poison reverse simply indicates that the learned route is unreachable. It is more reliable than split horizon alone. Examine Figure 3-7. Instead of suppressing the route for Network D, Router 2 sends that route in the routing table marked as unreachable. In RIP, the poison-reverse route is marked with a metric of 16 (infinite) to prevent that path from being used.

Triggered Updates

Another loop-prevention and fast-convergence technique used by routing protocols is triggered updates. When a router interface changes state (up or down), the router is required to send an update message, even if it is not time for the periodic update message. Immediate notification about a network outage is key to maintaining valid routing entries in all routers in the network by allowing faster convergence. Some distance-vector protocols, including RIP, specify a small delay to avoid having triggered updates generate excessive network traffic. The time delay is variable for each router.

Summarization

Another characteristic of routing protocols is the ability to summarize routes. Protocols that support VLSM can perform summarization outside IP class boundaries. By summarizing, a routing protocol can reduce the size of the routing table, and fewer routing updates occur on the network.

EIGRP Components

EIGRP is characterized by four components:

• Protocol-dependent modules

• Neighbor discovery and recovery

• Reliable Transport Protocol (RTP)

• Diffusing Update Algorithm (DUAL)

You should know the role of the EIGRP components, which are described in the following sections.

Router# show ip eigrp neighbor
IP-EIGRP neighbors for process 100
H  Address         Interface   Hold Uptime SRTT RTO Q    Seq Type
                   c           (sec)       (ms)     Cnt  Num
1 172.17.1.1      Se0         11 00:11:27   16  200  0   2
0 172.17.2.1      Et0         12 00:16:11   22  200  0   3

Router8# show ip eigrp topology
IP-EIGRP Topology Table for AS(100)/ID(172.16.3.1)


Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,
        r - reply Status, s - sia Status


P 172.16.4.0/30, 1 successors, FD is 2195456
           via 172.16.1.1 (2195456/2169856), Ethernet0
           via 172.16.5.1 (2376193/2348271), Ethernet1
P 172.16.1.0/24, 1 successors, FD is 281600
           via Connected, Ethernet0

EIGRP Timers

EIGRP sends updates only when necessary and sends them only to neighboring routers. There is no periodic update timer.

EIGRP uses hello packets to learn of neighboring routers. On high-speed networks, the default hello packet interval is 5 seconds. On multipoint networks with link speeds of T1 and slower, hello packets are unicast every 60 seconds.

The holdtime to maintain a neighbor adjacency is three times the hello time: 15 seconds. If a router does not receive a hello within the holdtime, it removes the neighbor from the table. Hellos are multicast every 60 seconds on multipoint WAN interfaces with speeds less than 1.544 Mbps, inclusive. The neighbor holdtime is 180 seconds on these types of interfaces. To summarize, hello/holdtime timers are 5/15 seconds for high-speed links and 60/180 seconds for multipoint WAN links less than 1.544 Mbps, inclusive.

EIGRP Metrics

EIGRP uses the same composite metric as IGRP, but the bandwidth (BW) term is multiplied by 256 for finer granularity. The composite metric is based on bandwidth, delay, load, and reliability. MTU is not an attribute for calculating the composite metric.

EIGRP calculates the composite metric with the following formula:

EIGRP_metric = {k1 × BW + [(k2 × BW)/(256 – load)] + k3 × delay} × {k5/(reliability + k4)}

In this formula, BW is the lowest interface bandwidth in the path, and delay is the sum of all outbound interface delays in the path. The router dynamically measures reliability and load. It expresses 100% reliability as 255/255. It expresses load as a fraction of 255. An interface with no load is represented as 1/255.

Bandwidth is the inverse minimum bandwidth (in kbps) of the path, in bits per second, scaled by a factor of 256 × 10⁷. The formula for bandwidth is

(256 × 10⁷)/BW_min

The delay is the sum of the outgoing interface delays (in tens of microseconds) to the destination. A delay of all 1s (that is, a delay of hexadecimal FFFFFFFF) indicates that the network is unreachable. The formula for delay is

sum_of_delays × 256

Reliability is a value between 1 and 255. Cisco IOS routers display reliability as a fraction of 255. That is, 255/255 is 100% reliability, or a perfectly stable link; a value of 229/255 represents a 90% reliable link.

Load is a value between 1 and 255. A load of 255/255 indicates a completely saturated link. A load of 127/255 represents a 50% saturated link.

By default, k1 = k3 = 1 and k2 = k4 = k5 = 0. EIGRP’s default composite metric, adjusted for scaling factors, is

EIGRP_metric = 256 × { [10⁷/BW_min] + [sum_of_delays] }

BW_min is in kbps, and sum_of_delays is in tens of microseconds. The bandwidth and delay for an Ethernet interface are 10 Mbps and 1 ms, respectively.

The calculated EIGRP BW metric is

256 × 10⁷/BW = 256 × 10⁷/10,000

= 256 × 1000

= 256,000

The calculated EIGRP delay metric is

256 × sum of delay = 256 × 1 ms

= 256 × 1000 × 1 microseconds

= 256,000 microseconds

Table 3-9 shows some default values for bandwidth and delay.

The metric weights subcommand is used to change EIGRP metric computation. You can change the k values in the EIGRP composite metric formula to select which EIGRP metrics to use. The command to change the k values is the metric weights tos k1 k2 k3 k4 k5 subcommand under router eigrp n. The tos value is always 0. You set the other arguments to 1 or 0 to alter the composite metric. For example, if you want the EIGRP composite metric to use all the parameters, the command is as follows:

router eigrp n
 metric weights 0 1 1 1 1 1

EIGRP Packet Types

EIGRP uses five packet types:

• Hello: EIGRP uses hello packets in the discovery of neighbors. They are multicast to 224.0.0.10. By default, EIGRP sends hello packets every 5 seconds (60 seconds on WAN links with 1.544 Mbps speeds or less).

• Acknowledgment: An acknowledgment packet acknowledges the receipt of an update packet. It is a hello packet with no data. EIGRP sends acknowledgment packets to the unicast address of the sender of the update packet.

• Update: Update packets contain routing information for destinations. EIGRP unicasts update packets to newly discovered neighbors; otherwise, it multicasts update packets to 224.0.0.10 when a link or metric changes. Update packets are acknowledged to ensure reliable transmission.

• Query: EIGRP sends query packets to find feasible successors to a destination. Query packets are always multicast unless they are sent as a response; then they are unicast back to the originator.

• Reply: EIGRP sends reply packets to respond to query packets. Reply packets provide a feasible successor to the sender of the query. Reply packets are unicast to the sender of the query packet.

EIGRP Design

When designing a network with EIGRP, remember that it supports VLSM and network summarization. EIGRP allows for the summarization of routes in a hierarchical network. EIGRP is not limited to 16 hops as RIP is; therefore, the network diameter can exceed this limit. In fact, the EIGRP diameter can be 225 hops. The default diameter is 100. EIGRP can be used in site-to-site WANs and IPsec virtual private networks (VPNs). In the enterprise campus, EIGRP can be used in data centers, server distribution, building distribution, and the network core.

EIGRP does not broadcast its routing table periodically, so there is no large network overhead. You can use EIGRP for large networks; it is a potential routing protocol for the core of a large network. EIGRP further provides for route authentication.

As shown in Figure 3-8, when you use EIGRP, all segments can have different subnet masks.

EIGRP is suited for almost all enterprise environments, including LANs and WANs, and is simple to design. The only caveat is that it is a Cisco-proprietary routing protocol that cannot be used with routers from other vendors. The use of EIGRP is preferred over RIP in all environments.

EIGRP Stub Routers

EIGRP allows for the configuration of stub routers for remote branches. It is used to reduce EIGRP query traffic between hub routers and remote branch routers that are connected over WAN links. EIGRP stub routing conserves memory and CPU resources and improves network stability. When the stub routing feature is enabled on the spoke router, the router only advertises specified routes to the hub router. The router does not advertise routes received from other EIGRP neighbors to the hub router. The only disadvantage is that the stub router cannot be used as a backup path between two hub sites.

Figure 3-9 shows an example of EIGRP stub router operation. If the LAN 10.10.10.0/24 goes down, the Hub1 router sends query packets everywhere; however, there is no need to send query packets to stub branches because there are no alternate routes there. Once you configure the branch routers as EIGRP stub routers, the query is sent only to the Hub2 router.

There are a few options when configuring the EIGRP stub routers:

• Receive-only: The stub router does not advertise network.

• Connected: The stub router can advertise directly connected networks.

• Static: The stub router can advertise static routes.

• Summary: The stub router can advertise summary routes.

• Redistribute: The stub router can advertise redistributed routes.

EIGRP for IPv4 Summary

The characteristics of EIGRP for IPv4 are as follows:

• EIGRP for IPv4 is a hybrid routing protocol (a distance-vector protocol that has link-state protocol characteristics).

• EIGRP for IPv4 uses IP protocol number 88.

• EIGRP for IPv4 is a classless protocol; it supports VLSM.

• The default composite metric uses bandwidth and delay.

• You can factor load and reliability into the metric.

• EIGRP for IPv4 sends partial route updates only when there are changes.

• EIGRP for IPv4 supports MD5 authentication.

• EIGRP for IPv4 uses DUAL for loop prevention.

• EIGRP for IPv4 provides fast convergence.

• By default, EIGRP for IPv4 uses equal-cost load balancing with equal metrics. Unequal-cost load sharing is possible with the variance command.

• The administrative distances are 90 for EIGRP internal routes, 170 for EIGRP external routes, and 5 for EIGRP summary routes.

• EIGRP for IPv4 allows for good scalability and is used in large networks.

• EIGRP for IPv4 multicasts updates to 224.0.0.10.

• EIGRP for IPv4 does not require a hierarchical physical topology.

EIGRP for IPv6 (EIGRPv6) Networks

EIGRP was originally an IPv4 routing protocol, although Cisco has developed IPv6 support into EIGRP to route IPv6 prefixes. EIGRP for IPv6 is configured and managed separately from EIGRP for IPv4; no network statements are used. EIGRP for IPv6 retains all the same characteristics (network discovery, DUAL, modules) and functions as EIGRP for IPv4. The major themes with EIGRP for IPv6 are as follows:

• EIGRP for IPv6 implements protocol-independent modules.

• EIGRP for IPv6 allows for neighbor discovery and recovery.

• EIGRP for IPv6 provides reliable transport.

• It implements the DUAL algorithm for a loop-free topology.

• EIGRP for IPv6 uses the same metrics as EIGRP for IPv4 networks.

• EIGRP for IPv6 uses the same timers as EIGRP for IPv4.

• EIGRP for IPv6 uses the same concepts of feasible successors and feasible distance as EIGRP for IPv4.

• EIGRP for IPv6 uses the same packet types as EIGRP for IPv4.

• EIGRP for IPv6 is managed and configured separately from EIGRP for IPv4.

• EIGRP for IPv6 requires a router ID before it can start running.

• EIGRP for IPv6 is configured on interfaces. No network statements are used.

One difference between EIGRP for IPv4 and EIGRP for IPv6 is the use of IPv6 prefixes and the use of IPv6 multicast group FF02::A for EIGRP updates, which are sourced from the link-local IPv6 address. This means that neighbors do not need to share the same global prefix, except for those neighbors that are explicitly specified for unicast updates.

Another difference is that EIGRP for IPv6 defaults to a shutdown state for the routing protocols and must be manually or explicitly enabled on an interface to become operational. Because EIGRP for IPv6 uses the same characteristics and functions as EIGRP for IPv4, as covered in the previous section, they are not repeated here.

EIGRP for IPv6 Summary

The characteristics of EIGRP for IPv6 are as follows:

• EIGRP for IPv6 uses the same characteristics and functions as EIGRP for IPv4.

• EIGRP for IPv6 is a hybrid routing protocol (a distance-vector protocol that has link-state protocol characteristics).

• EIGRP for IPv6 uses Next Header protocol 88.

• EIGRP for IPv6 routes IPv6 prefixes.

• The default composite metric uses bandwidth and delay.

• You can factor load and reliability into the metric.

• EIGRP for IPv6 sends partial route updates only when there are changes.

• EIGRP for IPv6 supports MD5 authentication.

• EIGRP for IPv6 uses DUAL for loop prevention and fast convergence.

• By default, EIGRP for IPv6 uses equal-cost load balancing. Unequal-cost load balancing is possible with the variance command.

• The administrative distances are 90 for EIGRP internal routes, 170 for EIGRP external routes, and 5 for EIGRP summary routes.

• EIGRP for IPv6 uses IPv6 multicast FF02::A for EIGRP updates.

• EIGRP for IPv6 enables high scalability and can be used in large networks.

A CCNP should understand EIGRP-specific characteristics and benefits. Table 3-10 provides a summary for reference.

Characteristic	EIGRP Support
Distance vector or link state	Hybrid: distance-vector routing protocol with link-state characteristics
Convergence	Fastest convergence with DUAL for a loop-free topology
Classless or classful	Classless routing protocol, supports VLSM
Scalability	Highly scalable, supports large networks
Multiprotocol support	Supports IPv4, IPv6, and legacy protocols such as IPX and AppleTalk
Multicast address for updates	224.0.0.10 for IPv4; FF02::A for IPv6

IS-IS Metrics

IS-IS, as originally defined, uses a composite metric with a maximum path value of 1024. The required default metric is arbitrary and is typically assigned by a network administrator. By convention, it is intended to measure the capacity of the circuit for handling traffic, such as its throughput in bits per second. Higher values indicate lower capacity. Any single link can have a maximum value of 64. IS-IS calculates path values by summing link values. The standard sets the maximum metric values to provide the granularity to support various link types, while ensuring that the shortest path algorithm used for route computation is reasonably efficient.

In Cisco routers, all active interfaces have a default metric of 10. If an interface is passive, the default value is 0. The administrator must configure the interface metric to get a different value. This small metric value range has proved insufficient for large networks and provides too little granularity for new features such as traffic engineering and other applications, especially with high-bandwidth links. Cisco IOS software addresses this issue with the support of a 24-bit metric field, the “wide metric.” Wide metrics are also required for route leaking. Using the new metric style, link metrics now have a maximum value of 16,777,215 (2²⁴ – 1) with a total path metric of 4,261,412,864 (254 × 2²⁴ = 2³²). Deploying IS-IS in the IP network with wide metrics is recommended for enabling finer granularity and supporting future applications such as traffic engineering.

IS-IS also defines three optional metrics (costs): delay, expense, and error. Cisco routers do not support the three optional metrics. The wide metric noted earlier uses the octets reserved for these metrics.

IS-IS Operation and Design

This section discusses IS-IS areas, designated routers, authentication, and the Network Entity Title (NET). IS-IS defines areas differently from OSPF; area boundaries are links and not routers. IS-IS has no BDRs. Because IS-IS is an OSI protocol, it uses a NET to identify each router.

IS-IS NET Addressing

Although you can configure IS-IS to route IP, the communication between routers uses OSI PDUs. The NET is the OSI address used for each router to communicate with OSI PDUs. A NET address ranges from 8 to 20 bytes in length and is hexadecimal. It consists of an authority and format identifier (AFI), area ID, system ID, and selector (SEL), as shown in Figure 3-10. The system ID must be unique within the network. An example of an IS-IS NET is 49.0001.1290.6600.1001.00, which consists of the following parts:

• AFI: 49

• Area ID: 0001

• System ID: 1290.6600.1001

• SEL: 00

Level 2 routers use the area ID. The system ID must be the same length for all routers in an area. For Cisco routers, it must be 6 bytes in length. Usually, a router MAC address identifies each unique router. The SEL is configured as 00. You configure the NET with the router isis command. In this example, the domain authority and format identifier (AFI) is 49, the area is 0001, the system ID is 00aa.0101.0001, and the SEL is 00:

router isis
net 49.0001.00aa.0101.0001.00

IS-IS Area Design

IS-IS uses a two-level hierarchy. Routers are configured to route Level 1 (L1), Level 2 (L2), or both Level 1 and Level 2 (L1/L2). Level 1 routers are like OSPF internal routers in a Cisco totally stubby area. (OSPF is covered in Chapter 4.) An L2 router is similar to an OSPF backbone router. A router that has both Level 1 and 2 routes is similar to an OSPF area border router (ABR). IS-IS does not define a backbone area like OSPF’s area 0, but you can consider the IS-IS backbone a continuous path of adjacencies among Level 2 ISs. When creating a backbone, there should never be Level 1 routers between L2-only or L1/L2 routers.

L1/L2 routers maintain separate link-state databases for the L1 routes and for the L2 routes. Also, the L1/L2 routers do not advertise L2 routes to the L1 area. L1 routers do not have information about destinations outside the area and use L1 routes to the L1/L2 routers to reach outside destinations.

As shown in Figure 3-11, IS-IS areas are not bounded by the L1/L2 routers but by the links between L1/L2 routers and L2 backbone routers.

With IS-IS, you can create three types of topologies:

• Flat

• Hierarchical

• Hybrid

The IS-IS flat design consists of a single area. As shown in Figure 3-12, all routers are on the same level. There is no summarization with this design. This design is recommended for a small network.

An IS-IS hierarchical topology is recommended for large networks. A backbone area can be created, much as with OSPF, and can have other areas connected via L1/L2 routers, as shown in Figure 3-13. This design allows for network summarization.

A hybrid IS-IS design is recommended for networks that might not be large enough for a backbone area. As shown in Figure 3-14, the IS-IS areas are connected using L1/L2 routers. This design allows for summarization between areas.

IS-IS Summary

The characteristics of IS-IS are as follows:

• It is a link-state protocol.

• It uses OSI CNLP to communicate with routers.

• It is a classless protocol (and supports VLSM and CIDR).

• The default metric is set to 10 for all active interfaces.

• IS-IS has two interface types: point-to-point and broadcast.

• It uses a single path metric, with a single link maximum of 64 and a path maximum of 1024.

• It sends partial route updates only when there are changes.

• IS-IS authentication uses plaintext passwords.

• The administrative distance is 115.

• It is used in large networks and is sometimes attractive as compared to OSPF and EIGRP.

• It is described in ISO/IEC 10589, reprinted by the IETF as RFC 1142.

• IS-IS provides support for IPv4 and IPv6 as separate topologies.