Metrics, DUAL, and the Topology Table

When implementing EIGRP, it is important to understand the various aspects used within and by the protocol before routes are actually placed into the IP routing table. In this section, you will learn about the EIGRP composite metric and how it is calculated. You will also learn about the different ways to influence metric calculation, as well as to adjust the calculated metric.

Following that, you will learn about the Diffusing Update Algorithm (DUAL) and the EIGRP topology table. This section concludes with a discussion on how all this information meshes when it comes to populating the IP routing table on a router running EIGRP.

EIGRP Composite Metric Calculation

Enhanced IGRP uses a composite metric, which includes different variables referred to as the K values. The K values are constants that are used to distribute weight to different path aspects, which may be included in the composite EIGRP metric. The default values for the K values are K1 = K3 = 1 and K2 = K4 = K5 = 0. In other words, K1 and K3 are set to a default value of 1, while K2, K4, and K5 are set to a default value of 0.

Assuming the default K value settings, the complete EIGRP metric can be calculated using the following mathematical formula:

However, given that only K1 and K3 have any positive values by default, the default EIGRP metric calculation is performed using the following mathematical formula:

This essentially means that, by default, EIGRP uses the minimum bandwidth on the path to a destination network and the total cumulative delay to compute routing metrics. However, Cisco IOS software allows administrators to set other K values to non-zero values to incorporate other variables into the composite metric. This may be performed using the metric weights [tos] k1 k2 k3 k4 k5 router configuration command.

When using the metric weights command, [tos] stands for Type of Service. Although Cisco IOS software shows that any value between 0 and 8 may be used, as of the time this guide was written, this field can currently be set only to zero. The K values can be set to any value between 0 and 255. The default EIGRP K values can be viewed by issuing the show ip protocols command. This is illustrated in the following output:

When adjusting the EIGRP K values, it is important to remember that the same values must be configured on all routers within the EIGRP domain. If the K values are mismatched, EIGRP neighbour relationships will not be established.

NOTE: Adjusting the default K value settings is not recommended. It should be done only with the assistance of seasoned senior-level engineers who have a solid understanding of the implications of such actions within the network or based upon the recommendation of the Cisco Technical Assistance Centre (TAC).

Using Interface Bandwidth to Influence EIGRP Metric Calculation

Enhanced IGRP metric calculation can be directly influenced by adjusting the default bandwidth values assigned to individual interfaces using the bandwidth command. The bandwidth values specified by this command are in Kilobits. The bandwidth used in EIGRP metric calculation is also in Kilobits. Figure 36.8 below illustrates a network comprised of two routers connected via two Serial (T1) links that have a bandwidth value of 1544Kbps:

Figure 36.8 – EIGRP Metric Bandwidth Manipulation

Referencing the diagram in Figure 36.8, because of the equal bandwidth (and delay) values of the links between R1 and R2, the same EIGRP metric will be derived for both paths from R2 to the 172.16.100.0/24 subnet. EIGRP will load-share traffic between the two Serial links, as illustrated in the following output on R2:

Adjusting the default bandwidth value on either interface will directly influence the EIGRP metric calculation for the path to the destination network. Such actions can be used for path control within larger networks (i.e., controlling the path that traffic takes based on administrator-defined values and configurations). For example, if it was preferred that EIGRP use Serial0/0 as the primary path to the destination network and Serial0/1 as the backup path to the destination, one of two actions could be taken.

The first is that the bandwidth value on Serial0/0 could be incremented, resulting in a better (lower) metric for this path. The second is that the bandwidth value on Serial0/1 could be decremented, resulting in a worse (higher) metric for this path. Either option is acceptable and will achieve the desired result. The following output illustrates how to decrement the default bandwidth value on Serial0/1, effectively ensuring that Serial0/0 is used as the primary path between R2 and the 172.16.100.0/24 network:

NOTE: As stated in Day 1, this configuration does not mean that Serial0/1 is now capable of only 1024Kbps of throughput through this interface.

The result of this configuration is that Serial0/0 is the primary path used by R2 to get to the 172.16.100.0/24 destination network. This is illustrated in the following output:

NOTE: The asterisk (*) points to the interface over which the next packet is sent. In the event that there are multiple equal-cost routes in the routing table, the position of the * rotates among the equal-cost paths.

Although the path via the Serial0/1 interface is not installed into the routing table, when using EIGRP as the routing protocol, it is important to remember that this path is not completely ignored. Instead, this path is stored in the EIGRP topology table, which contains the primary and alternate (backup) paths to remote destination networks. The EIGRP topology table will be described in detail later in this module.

NOTE: By default, when EIGRP is enabled, it can use up to 50% of the interface bandwidth to send EIGRP packets (EIGRP is a very chatty protocol, so it limits itself in possible bandwidth usage). EIGRP determines the bandwidth amount based on the bandwidth interface configuration command. Therefore, when adjusting interface bandwidth values, it is important to keep this fact in mind. This default setting can be adjusted by using the ip bandwidth-percent eigrp [ASN] [percentage] interface configuration command.

In summation, when using the bandwidth command to influence EIGRP metric calculation, it is important to remember that EIGRP uses the minimum bandwidth on the path to a destination network, along with the cumulative delay, to compute routing metrics. It is important to have a solid understanding of the network topology to best determine where to use the bandwidth command to influence EIGRP metric calculation. In the real world, however, delay is the preferred method of influencing EIGRP metrics.

Using Interface Delay to Influence EIGRP Metric Calculation

The interface delay value is presented in microseconds. The delay value used in EIGRP metric calculation is in tens of microseconds. Therefore, the delay value on the interface must be divided by 10 in order to compute the EIGRP metric. Table 36.3 below shows the default interface bandwidth and delay values used in Cisco IOS software:

Table 36.3 – Default Interface Bandwidth and Delay Values

1000000

44736

34010

200

When working with the interface bandwidth and delay values, it is very important to remember that adjusting the interface bandwidth value does not automatically adjust the interface delay value, and vice-versa. The two values are independent of each other. As an example, the output that follows shows the default bandwidth and delay values for a FastEthernet interface:

To reinforce this concept, the bandwidth value on the FastEthernet interface is adjusted to 1544Kbps using the bandwidth interface configuration command, as follows:

While the bandwidth value now displayed in the output of the show interfaces command reflects the implemented configuration, the default interface delay value remains the same, as illustrated below:

R2#show interfaces FastEthernet0/0
FastEthernet0/0 is up, line protocol is up
Hardware is AmdFE, address is 0013.1986.0a20 (bia 0013.1986.0a20)
Internet address is 192.168.1.2/24
MTU 1500 bytes, BW 1544 Kbit/sec, DLY 100 usec,
reliability 255/255, txload 1/255, rxload 1/255

The cumulative delay used by EIGRP is the sum of all interface delays between the source and the destination network. Changing any of the delay values in the path influences EIGRP metric calculation. The interface delay value is adjusted using the delay interface configuration command. This value is then divided by 10 when used in EIGRP metric calculation. Figure 36.9 below illustrates a network comprised of two routers connected via two Serial (T1) links that have a bandwidth value of 1544Kbps and a default delay of 20000 microseconds. In addition, the 172.16.100.0/24 network is directly connected to a FastEthernet interface, which has a default bandwidth of 100000Kbps and a default delay value of 100 microseconds:

Figure 36.9 – EIGRP Metric Delay Manipulation

The EIGRP metric from R2 to the 172.16.100.0/24 network is calculated as follows:

Metric = [(10⁷ ⁄ least bandwidth on path) + (sum of all delays)] × 256

Metric = [(10000000 ⁄ 1544) + (2000 + 10)] × 256

NOTE: Remember to divide the interface delay values by 10 for EIGRP metric calculation.

Metric = [(10000000 ⁄ 1544) + (2000 + 10)] × 256

NOTE: The calculated value should always be rounded down to the nearest integer.

Metric = [6476 + 2010] × 256

Metric = 8486 × 256

Metric = 2172416

This calculation can be verified by the show ip route command, as follows:

As with the bandwidth command, you can either increment or decrement the interface delay value using the delay command to influence EIGRP metric calculation. For example, to configure R2 to use the Serial0/0 link to get to the 172.16.100.0/24 network, with Serial0/1 being used as a backup link only, the delay value on Serial0/0 could be decremented as follows:

This configuration adjusts the EIGRP metric for the path via Serial0/0, as illustrated below:

The path via Serial0/1 is retained in the topology table as an alternate path to the network.

The Diffusing Update Algorithm (DUAL)

The Diffusing Update Algorithm is at the crux of the EIGRP routing protocol. DUAL looks at all routes received from neighbour routers, compares them, and then selects the lowest metric (best) loop-free path to the destination network, which is the Feasible Distance (FD), resulting in the Successor route. The FD includes both the metric of a network as advertised by the connected neighbour plus the cost of reaching that particular neighbour.

The metric that is advertised by the neighbour router is referred to as the Reported Distance (RD) or as the Advertised Distance (AD) to the destination network. Therefore, the FD includes the RD plus the cost of reaching that particular neighbour. The next-hop router for the Successor route is referred to as the Successor. The Successor route is placed into the IP routing table and the EIGRP topology table and points to the Successor.

Any other routes to the same destination network that have a lower RD than the FD of the Successor path are guaranteed to be loop-free and are referred to as Feasible Successor (FS) routes. These routes are not placed into the IP routing table; however, they are still placed into the EIGRP topology table, along with the Successor routes.

In order for a route to become an FS route, it must meet the Feasibility Condition (FC), which occurs only when the RD to the destination network is less than the FD. In the event that the RD is more than the FD, the route is not selected as an FS. This is used by EIGRP to prevent the possibility of loops. The network topology illustrated in Figure 36.10 below will be used to clarify the terminology referred to in this section:

Figure 36.10. – Understanding the Diffusing Update Algorithm

Referencing Figure 36.10, Table 36.4 below shows the Feasible Distance and the Reported Distance values as seen on R1 for the 192.168.100.0/24 network:

Table 36.4 – R1 Paths and Distances

Based on the information in Table 36.4, R1 will select the path through R4 as the Successor route based on the FD for the route, which is 25. This route will be placed into the IP routing table as well as the EIGRP topology table. R1 then looks at alternate paths to the 192.168.100.0/24 network. The metric for neighbour R3 to the 192.168.100.0/24 network, also referred to as the RD or AD, is 10. This is less than the FD and so this route meets the FC and is placed into the EIGRP topology table. The metric for neighbour R2 to the 192.168.100.0/24 network is 30. This value is higher than the FD of 25. This route does not meet the FC and is not considered an FS. The route, however, is still placed into the EIGRP topology table. This is illustrated in the section on the EIGRP topology table that follows.

When a neighbour changes a metric, or when a topology change occurs, and the Successor route is removed or changes, DUAL checks for FSs for the route and if one is found, then DUAL uses it to avoid re-computing the route unnecessarily. This is referred to as local computation. Performing a local computation saves CPU power because the FS has been chosen and already exists before the Successor or primary route fails.

When no FS for the destination network exists, the local router will send a Query to neighbouring routers asking if they have information on the destination network. If the information is available and another neighbour does have a route to the destination network, then the router performs a diffusing computation to determine a new Successor.

The EIGRP Topology Table

The EIGRP topology table is populated by EIGRP PDMs acted upon by the DUAL Finite State Machine. All known destination networks and subnets that are advertised by neighbouring EIGRP routers are stored in the EIGRP topology table. This includes Successor routes, FS routes, and even routes that have not met the FC.

The topology table allows all EIGRP routers to have a consistent view of the entire network. It also allows for rapid convergence in EIGRP networks. Each individual entry in the topology table contains the destination network and the neighbour(s) that have advertised the destination network. Both the FD and the RD are stored in the topology table. The EIGRP topology table contains the information needed to build a set of distances and vectors to each reachable network, including the following:

• The lowest bandwidth on the path to the destination network
• The total or cumulative delay to the destination network
• The reliability of the path to the destination network
• The loading of the path to the destination network
• The minimum Maximum Transmission Unit (MTU) to the destination network
• The Feasible Distance to the destination network
• The Reported Distance by the neighbour router to the destination network
• The route source (only external routes) of the destination network

NOTE: While the MTU is included in the topology table, EIGRP does not use this value in actual metric computation. Instead, the MTU is simply tracked to determine the minimum value to the destination network. The interface MTU specifies the largest size of datagram that can be transferred across a certain link without the need for fragmentation, or breaking the datagram or packet into smaller pieces.

The contents of the EIGRP topology table are viewed using the show ip eigrp topology command. The options that are available with this command are illustrated below:

The show ip eigrp topology command with no options prints only the Successor and Feasible Successor information for routes in the topology table and for all of the EIGRP instances enabled on the router. The output printed by this command is illustrated below:

The show ip eigrp topology [network]/[prefix] and show ip eigrp topology [network] [mask] commands print Successor routes, FS routes, and routes that have not met the FC for the route specified in either command. The following illustrates the use of the show ip eigrp topology [network]/[prefix]command:

In the output above, you can see that the path via Serial0/1 does not meet the FC because the RD exceeds the FD. This is why the path is not printed in the output of the show ip eigrp topology command. Instead of viewing each prefix on an individual basis to determine Successor routes, FS routes, and routes that did not meet the FC, you can use the show ip eigrp topology all-links command to view all possible routes for all of the prefixes in the EIGRP topology table. The output of this command is illustrated below:

Within the EIGRP topology table, entries may be marked either as Passive (P) or as Active (A). A route in the Passive state indicates that EIGRP has completed actively computing the metric for the route and traffic can be forwarded to the destination network using the Successor route. This is the preferred state for all routes in the topology table.

Enhanced IGRP routes are in an Active state when the Successor route has been lost and the router sends out a Query packet to determine an FS. Usually, an FS is present and EIGRP promotes that to the Successor route. This way, the router converges without involving other routers in the network. This process is referred to as a local computation.

However, if the Successor route has been lost or removed, and there is no FS, then the router will begin diffused computation. In diffused computation, EIGRP will send a Query out to all neighbours and out of all interfaces, except for the interface to the Successor route. When an EIGRP neighbour receives a Query for a route, and if that neighbour’s EIGRP topology table does not contain an entry for the route, then the neighbour immediately replies to the Query with an unreachable message, stating that there is no path for this route through this neighbour.

If the EIGRP topology table on the neighbour lists the router sending the Query as the Successor for that route, and an FS exists, then the FS is installed and the router replies to the neighbour Query that it has a route to the lost destination network.

However, if the EIGRP topology table lists the router sending the Query as the Successor for this route and there is no FS, then the router queries all of its EIGRP neighbours, except those that were sent out of the same interface as its former Successor. The router will not reply to the Query until it has received a Reply to all Queries that it originated for this route.

Finally, if the Query was received from a neighbour that is not the Successor for this destination, then the router replies with its own Successor information. If the neighbouring routers do not have the lost route information, then Queries are sent from those neighbouring routers to their neighbouring routers until the Query boundary is reached. The Query boundary is either the end of the network, the distribute list boundary, or the summarisation boundary.

Once the Query has been sent, the EIGRP router must wait for all replies to be received before it calculates the Successor route. If any neighbour has not replied within three minutes, the route is said to be Stuck-in-Active (SIA). When a route is SIA, the neighbour relationship of the router(s) that did not respond to the Query will be reset. In such cases, you will see a message logged by the router similar to the following:

There are several reasons why the EIGRP neighbour router(s) may not respond to the Query, which include the following:

• The neighbour router’s CPU is overloaded and it cannot respond in time
• The neighbour router itself has no information about the lost route
• Quality issues on the circuit are causing packets to be lost
• Low-bandwidth links are congested and packets are being delayed

To prevent SIA issues due to delayed responses from other EIGRP neighbours, the local router can be configured to wait for longer than the default of three minutes to receive responses back to its Query packets using the timers active-time command in Router Configuration mode.

NOTE: It is important to note that if you change this default parameter on one EIGRP router in your network, you must change it on all the other routers within your EIGRP routing domain.