EIGRP Concepts and Operation

November 10, 2016 by Neel Rao

Filed under Network

Last modified November 10, 2016

EIGRP Concepts and Operation

Like OSPF, EIGRP follows three general steps to be able to add routes to the IP routing table:

  • Neighbor discovery: EIGRP routers send Hello messages to discover potential neighboring EIGRP routers and perform basic parameter checks to determine which routers should become neighbors.
  • Topology exchange: Neighbors exchange full topology updates when the neighbor relationship comes up, and then only partial updates as needed based on changes to the network topology.
  • Choosing routes: Each router analyzes its respective EIGRP topology tables, choosing the lowest-metric route to reach each subnet.

As a result of these three steps, IOS maintains three important EIGRP tables. The EIGRP neighbor table lists the neighboring routers and is viewed with the show ip eigrp neighbor command. The EIGRP topology table holds all the topology information learned from EIGRP neighbors and is  displayed with the show ip eigrp topology command. Finally, the IP routing table holds all the best routes and is displayed with the show ip route command.

The next few sections describe some details about how EIGRP forms neighbor relationships, exchanges routes, and adds entries to the IP routing table. In addition to these three steps, this section explains some unique logic EIGRP uses when converging and reacting to changes in an internetwork—logic that is not seen with the other types of routing protocols.

EIGRP Neighbors

An EIGRP neighbor is another EIGRP-speaking router, connected to a common subnet, with which the router is willing to exchange EIGRP topology information. EIGRP uses EIGRP Hello messages, sent to multicast IP address 224.0.0.10, to dynamically discover potential neighbors. A router learns of potential neighbors by receiving a Hello.

Routers perform some basic checking of each potential neighbor before that router becomes an EIGRP neighbor. A potential neighbor is a router from which an EIGRP Hello has been received. Then the router checks the following settings to determine if the router should be allowed to be a neighbor:

  • It must pass the authentication process.
  • It must use the same configured AS number.
  • The source IP address used by the neighbor’s Hello must be in the same subnet.

The verification checks are relatively straightforward. If authentication is configured, the two routers must be using the same type of authentication and the same authentication key.

EIGRP configuration includes a parameter called an autonomous system number (ASN), which must be the same on two neighboring routers. Finally, the IP addresses used to send the EIGRP Hello messages—the routers’ respective interface IP addresses—must be in the range of addresses on the other routers’ respective connected subnet.

The EIGRP neighbor relationship is much simpler than OSPF. EIGRP does not have an additional concept of being fully adjacent like OSPF, and there are no neighbor states like OSPF. As soon as an EIGRP neighbor is discovered and passes the basic verification checks, the router becomes a neighbor. At that point, the two routers can begin exchanging topology information. The neighbors send Hellos every EIGRP Hello interval. A router considers its EIGRP neighbor to no longer be reachable after the neighbor’s Hellos cease to occur for the number of seconds defined by the EIGRP Hold Timer—the rough equivalent of the OSPF Dead Interval.

Exchanging EIGRP Topology Information

EIGRP uses EIGRP Update messages to send topology information to neighbors. These Update messages can be sent to multicast IP address 224.0.0.10 if the sending router needs to update multiple routers on the same subnet; otherwise, the updates are sent to the unicast IP address of the particular neighbor. (Hello messages are always sent to the 224.0.0.10 multicast address.) Unlike OSPF, there is no concept of a Designated Router (DR) or Backup Designated Router (BDR), but the use of multicast packets on LANs allows EIGRP to exchange routing information with all neighbors on the LAN efficiently.

The update messages are sent using Reliable Transport Protocol (RTP). The significance of RTP is that, like OSPF, EIGRP resends routing updates that are lost in transit. By using RTP, EIGRP can better avoid loops.

Neighbors use both full routing updates and partial updates, as shown in Figure 10-1. A full update means that a router sends information about all known routes, whereas a partial update includes only information about recently changed routes. Full updates occur when neighbors first come up. After that, the neighbors send only partial updates in reaction to changes to a route. From top to bottom, Figure 10-1 shows neighbor discovery with Hellos, the sending of full updates, the maintenance of the neighbor relationship with ongoing Hellos, and partial updates.

EIGRP1

EIGRP1

Calculating the Best Routes for the Routing Table

 Metric calculation is one of the more interesting features of EIGRP. EIGRP uses a composite metric, calculated as a function of bandwidth and delay by default. The calculation can also include interface load and interface reliability, although Cisco recommends against using either. EIGRP calculates the metric for each possible route by inserting the values of the composite metric into a formula.

EIGRP’s metric calculation formula actually helps describe some of the key points about the metric. The formula, assuming that the default settings use just bandwidth and delay, is as follows:

EIGRP2

EIGRP2

In this formula, the term least-bandwidth represents the lowest-bandwidth link in the route, using a unit of kilobits per second. For instance, if the slowest link in a route is a 10-Mbps Ethernet link, the first part of the formula is 107 / 104, which equals 1000. You use 104 in the formula because 10 Mbps is equal to 10,000 kbps (104 kbps). The cumulative-delay value used in the formula is the sum of all the delay values for all links in the route, with a unit of “tens of microseconds.” You can set both bandwidth and delay for each link, using the cleverly named bandwidth and delay interface subcommands.

EIGRP updates list the subnet number and mask, along with the cumulative delay, minimum bandwidth, along with the other typically unused portions of the composite metric. The router then considers the bandwidth and delay settings on the interface on which the update was received and calculates a new metric. For example, Figure 10-2 shows Albuquerque learning about subnet 10.1.3.0/24 from Seville. The update lists a minimum bandwidth of 100,000 kbps, and a cumulative delay of 100 microseconds. R1 has an interface bandwidth set to 1544 kbps—the default bandwidth on a serial link—and a delay of 20,000 microseconds.

Figure 10-2 How Albuquerque Calculates Its EIGRP Metric for 10.1.3.0/24 NOTE Most show commands, including show ip eigrp topology and show interfaces, list delay settings as the number of microseconds of delay. The metric formula uses a unit of tens of microseconds.

EIGRP updates list the subnet number and mask, along with the cumulative delay, minimum bandwidth, along with the other typically unused portions of the composite metric. The router then considers the bandwidth and delay settings on the interface on which the update was received and calculates a new metric. For example, Figure 10-2 shows Albuquerque learning about subnet 10.1.3.0/24 from Seville. The update lists a minimum bandwidth of 100,000 kbps, and a cumulative delay of 100 microseconds. R1 has an interface bandwidth set to 1544 kbps—the default bandwidth on a serial link—and a delay of 20,000 microseconds.

EIGRP3

EIGRP3

In this case, Albuquerque discovers that its S0/1 interface bandwidth (1544) is less than the advertised minimum bandwidth of 100,000, so Albuquerque uses this new, slower bandwidth in the metric calculation. (If Albuquerque’s S0/1 interface had a bandwidth of 100,000 or more in this case, Albuquerque would instead use the minimum bandwidth listed in the EIGRP Update from Seville.) Albuquerque also adds the interface S0/1 delay (20,000 microseconds, converted to 2000 tens-of-microseconds for the formula) to the cumulative delay received from Seville in the update (100 microseconds, converted to 10 tens-of-microseconds). This results in the following metric calculation:

eigrp4

EIGRP4

NOTE IOS rounds down the division in this formula to the nearest integer before performing the rest of the formula. In this case, 107 / 1544 is rounded down to 6476.

If multiple possible routes to subnet 10.1.3.0/24 existed, Albuquerque would also calculate the metric for those routes and would choose the route with the best (lowest) metric to be added to the routing table. If the metric is a tie, by default a router would place up to four equal-metric routes into the routing table, sending some traffic over each route. The later section “EIGRP Maximum Paths and Variance” explains a few more details about how EIGRP can add multiple equal-metric routes, and multiple unequal-metric routes, to the routing table.

Feasible Distance and Reported Distance

The example described for Figure 10-2 provides a convenient backdrop to define a couple of EIGRP terms:

  • Feasible Distance (FD): The metric of the best route to reach a subnet, as calculated on a router.
  • Reported Distance (RD): The metric as calculated on a neighboring router and then reported and learned in an EIGRP Update

For example, in Figure 10-2, Albuquerque calculates an FD of 2,195,631 to reach subnet 10.1.3.0/24 through Seville. Seville also calculates its own metric to reach subnet 10.1.3.0/24. Seville also lists that metric in its EIGRP update sent to Albuquerque. In fact, based on the information in Figure 10-2, Seville’s FD to reach subnet 10.1.3.0/24, which is then known by Albuquerque as Seville’s RD to reach 10.1.3.0/24, could be easily calculated:

EIGRP5

EIGRP5

FD and RD are mentioned in an upcoming discussion of how EIGRP reacts and converges when a change occurs in an internetwork.

Caveats with Bandwidth on Serial Links

EIGRP’s robust metric gives it the ability to choose routes that include more router hops but with faster links. However, to ensure that the right routes are chosen, engineers must take care to configure meaningful bandwidth and delay settings. In particular, serial links default to a bandwidth of 1544 and a delay of 20,000 microseconds, as used in the example shown in Figure 10-2. However, IOS cannot automatically change the bandwidth and delay settings based on the Layer 1 speed of a serial link. So, using default bandwidth settings on serial links can lead to problems.

Figure 10-3 shows the problem with using default bandwidth settings and how EIGRP uses the better (faster) route when the bandwidth is set correctly. The figure focuses on router B’s route to subnet 10.1.1.0/24 in each case. In the top part of the figure, all serial interfaces use defaults, even though the top serial link is a slow 64 kbps. The bottom figure shows the results when the slow serial link’s bandwidth command is changed to reflect the correct (slow) speed.

EIGRP6

EIGRP6

 

Refer next blog : EIGRP Convergence

 

By N.R.Rao

For SkyBird Technology Solutions Pvt Ltd.

 

 

 

Leave a Comment