Balanced hybrid routing protocols are sometimes called advanced distance vector routing protocols. The rationale behind this logic is that these routing protocols use similar metrics and have a maximum hop count as distance vector routing protocols. However, balanced hybrid routing protocols discover neighbors and put them in a neighbor table before exchanging routing information, as well as keep lists of all possible routes in a topology table, just as link-state routing protocols do. By taking the best attributes from both classes of routing protocols, they have the pick of the litter, so to speak. This enables these routing protocols to be considered some of the more elite routing protocols.
EIGRP the biggest contender for routing protocol stardom is a Cisco-proprietary routing protocol called Enhanced Interior Gateway Routing Protocol. As the name states, EIGRP is an enhanced version of the (now-defunct) Cisco distance vector routing protocol, IGRP. This article looks at exactly how this routing protocol is actually one of the fastest converging protocols that exist today.
One of the most notable features that EIGRP offers is the use of a robust 32-bit composite metric. Specifically, EIGRP uses bandwidth and delay (each multiplied by 255 to make them 32-bit) as its default metrics to determine the best routing path to a destination. In addition, you can configure EIGRP to include additional metrics such as reliability, load, and MTU. By using this more robust composite metric (as opposed to using hop count or cost based on bandwidth), routers can accurately determine the best path to take a destination. For example, in the picture given below, RIPv2 would take the T1 link as the best route to reach 172.17.0.0, because it has the fewest hops.
EIGRP, on the other hand, considers the bandwidth of the link and determines that the bottom path is the optimal route, because the Fast Ethernet links combined are still faster than a single T1 or 1.54Mbps.
Because EIGRP metrics can vary considerably depending on the bandwidth and delay of the links to the destination, EIGRP also can support load balancing up to six unequal paths (as opposed to RIP’s and OSPF’s equal paths). What’s more, EIGRP can support larger networks than RIP, because its maximum hop count is 244 as compared to RIP and RIPv2’s 15 hops. EIGRP still requires a maximum hop count, because it shares some characteristics of distance vector routing protocols and uses mechanisms to avoid routing loops.
Because EIGRP is a true hybrid routing protocol, it has taken some of the best features of link-state routing protocols too. For instance, EIGRP discovers its neighbors and builds a topology table by sending hello messages as a multicast to the reserved multicast address of 220.127.116.11. after the neighbors are discovered, they synchronize their topology databases and send hello messages afterward to keep their dead timers from expiring. The time differs depending on the topology, just as you saw with OSPF. Specifically, point-to-point and broadcast topologies have a 5-second hello interval and 15-second dead timer, whereas nonbroadcast multi-access topologies such as Frame Relay have a 60-second hello interval and 180-second dead timer.
Cisco, however, did not stop there with the features of this balanced hybrid routing protocol. It developed a new routing algorithm called the Diffusing Update Algorithm (DUAL) that ensures a 100% loop-free routing environment that can converge in the face of a topology change in a split second. EIGRP also can route not only IP but also IPX and AppleTalk routed protocols in your network if you have an older Novell or Macintosh environment.
Another additional useful characteristics of EIGRP is its capability to distinguish between internally learned networks and networks that were redistributed into EIGRP. External networks get tagged when being redistributed, so EIGRP knows not to trust those networks over native EIGRP networks. EIGRP assigns the external networks an administrative distance (AD) of 170 and the internal networks an administrative distance of 90.
Last but not least, EIGRP is classful by default but can be configured to be classless, similar to RIPv2. By disabling automatic summarization, you can support VLSM designs and discontinuous networks, as well as manually summarize networks at any bit boundary you want.
The only significant downfall of this routing protocol is that all your routers must be Cisco routers (not that this is a bad thing) to support this proprietary routing protocol.
Successor and Feasible Successor Routes
The secret of EIGRP’s rapid convergence is found in its topology table. Just like OSPF, EIGRP stores all possible routes in the database and calculates the best path to each subnet, based on the lowest cumulative composite metric. Those best routes are known as the successor routes.
EIGRP keeps track of the composite metric for every subnet that is being advertised to it by neighbor routers, known as the advertised distance. The router also tracks that advertised distance plus the composite metric to reach that advertising router from the local router, known as the feasible distance. The lowest feasible distance to a particular subnet becomes the successor route and is the path that is also placed in the routing table.
Where EIGRP sets itself apart from OSPF is that it keeps an ace up its sleeve, so to speak. If the conditions are correct, EIGRP keeps a backup route in its topology table known as the feasible successor. In the event that a successor route fails, the feasible successor becomes the successor’s route and is placed in the routing table in about one second.
The feasible successor route is chosen only if the route will not cause a loop when activated and if the advertised distance from a neighbor is less than the existing successor route’s feasible distance. In other words, the feasible successor must have an advertised metric that is less than the metric of the route in the routing table. For example, to reach Network X, imagine your local router’s successor route might have a feasible distance of 8000. If any neighbor propagated an advertised distance for it to reach Network X of 7999 or less, its route is a feasible successor. If the advertised distances are 8000 or more, the route is in the topology table but is not a candidate for a feasible successor.
DUAL Algorithm in Action
You already know that the feasible successor route in the topology table is used if the primary successor route fails. So what happens if there isn’t a feasible successor, as illustrated in the picture below. In this picture, you see a glimpse of Router D’s topology table for the 172.17.0.0/16 subnet. The P next to the network stands for the passive state, which in EIGRP terms is actually a good thing.
Underneath, you see two possible routes through Router A and Router B with numbers in parentheses separated by a slash. The number on the left represents the advertised distance from the neighbor router, and the number on the right represents the feasible distance to reach that subnet through that advertising router. Because the path through Router B has the lowest feasible distance, that is the successor route, which is also placed in the routing table. The route through Router A has an advertised distance of 9700, which is not less than the successor route’s feasible distance (3700), so it cannot be a feasible successor.
Now let’s see what happens when the successor route to Router B fails, as shown in the picture below. Without a feasible successor route to 172.17.0.0, Router D puts that network into an active state. The network is active because the router begins to actively query its directly connected neighbors about whether they have a route to the affected network. This is considerably different from and less resource-intensive than OSPF, because the route only asks its neighbors, as opposed to flooding the update throughout the area.
When Router A responds to Router D’s query, it adds that entry to its topology table, which in turn becomes the new successor route and is placed in the routing table. To ensure a loop-free environment, Router D has to wait for all queries to come back before implementing the new route. That is why EIGRP routers start what is known as a Stuck in Active (SIA) timer, which is how long it waits for a response from a query. The default SIA timeout is 180 seconds.
EIGRP Stub Routing
It should come as no surprise that EIGRP borrowed another link-state concept to improve routing convergence and gave router resources. Similar to an OSPF stub area, EIGRP allows you to configure what is known as EIGRP stub routing. This configuration is most useful in instances where the network is designed in a hub-and-spoke topology, as shown in the picture below.
Because each of the remote routers (Routers B, C, and D) all have to pass data through the central Router A to reach the remotes, it does not make sense for EIGRP devices to act as they normally would when a topology change occurs. For example, if the 172.16.0.0 network connected to Router B goes down, what good does it do for the routers to go into an active state for that link and query its neighbors for an alternate route?
There is only one way in and out of that network (because that is the defining characteristic of a stub network). By configuring the remote routers as stub routers (the central, or hub, router does not need to be configured), the routers immediately respond to the query as “inaccessible” and speed up the time it takes to converge the network