| Objective: Manage TCP/IP routing. A routing protocol is based either on a distance-vector or link-state algorithm. The differences between the two algorithms include when routing information is exchanged, what information is sent when the routing information is exchanged, and how quickly the protocol can route around outages (when the network topology will support it). Distance-Vector Routing Distance-vector routing protocols require that each router tell its neighbors what is in its routing table. This is done by sending the entire routing table when the router boots and then sending it again at scheduled intervals. Each router uses updates from its neighboring routers to update its own routing table based on the information it receives. Using the information from these updates, a router can build a network map in its routing table and determine hop counts (that is, the distance to any network) for each network entry in the routing table. Routing updates sent using a distance-vector routing protocol are unacknowledged and unsynchronized, which is one of the drawbacks of distance-vector protocols. Some other drawbacks of this type of routing protocol include the following: High overhead Because every router on the network sends its entire routing table when it sends an update, distance-vector protocols produce very large routing tables. This adds overhead to the router memory needed to store these tables and the router processing power needed to maintain these tables. Large routing tables can also cause problems for an administrator who is trying to determine the source of a problem. Not scalable Distance-vector networks are limited to 15 hops for any given route. In a large network (such as the Internet), it is easy to have network segments that are greater than 15 hops away; these would be unreachable in a distance-vector network. Network bandwidth intensive Distance-vector networks require that routers exchange their entire routing tables whenever they do updates. On a large network with large routing tables, these routing updates can utilize significant amounts of bandwidth, especially across small WAN connections or demand-dial links. Long convergence time Convergence is the amount of time it takes a routing algorithm to detect and route around a network failure. Distance-vector protocols typically have longer convergence times than link-state protocols. Routing loops Distance-vector protocols can suffer from routing loop problems when there are multiple paths to a network. A routing loop occurs when the routing table creates a routing path that loops between routers. The packet is sent from router to router until its TTL is exceeded. Routing loops typically occur in poorly designed networks. Count-to-infinity problems Count-to-infinity problems occur when there is a network outage and the routing algorithm cannot calculate a new router. One router broadcasts a route and increments the hop count for the router, and then a second router broadcasts the same route to the first router, also incrementing the hop count, and so on, until the route metric (that is, the hop count) reaches 16 and the route is discarded.
Distance-vector routing protocols have additional mechanisms that allow them to avoid the count-to-infinity problems and to improve convergence. They include the following: Split horizon The split-horizon mechanism prevents routes from being broadcast out the interface from which they were received. Split horizon eliminates count-to-infinity problems and routing loops during convergence in single-path internetworks and reduces the chances of count-to-infinity problems occurring in multipath internetworks. Split horizon with poison reverse The split horizon with poison reverse mechanism allows routes to be broadcast back to the interface from which they were received, but they are announced with a hop count of 16, which indicates that the network is unreachable (in other words, the route has been poisoned and is unusable through that interface). In a single-path internetwork, split horizon with poison reverse has no benefits in addition to those of split horizon. However, in a multipath internetwork, split horizon with poison reverse greatly reduces routing loops and count-to-infinity problems, which can still occur in a multipath internetwork because routes to networks can be learned from multiple sources. Triggered updates Triggered updates allow a router to announce changes in metric values almost immediately, rather than wait for the next periodic announcement. The trigger is a change to a metric in an entry in the routing table. For example, networks that become unavailable can be announced with a hop count of 16 through a triggered update. If triggered updates were sent by all routers immediately, each triggered update could cause a cascade of broadcast traffic across the IP internetwork. Triggered updates improve the convergence time of RIP internetworks, but at the expense of additional broadcast traffic as the triggered updates are propagated.
The main advantages of distance-vector routing are that it requires little maintenance and that it is easy to configure. Both of these advantages make it popular in small network environments. Link-State Routing Link-state routing was designed to overcome the disadvantages of distance-vector routing, and it does a decent job of this. Routers using link-state routing protocols learn about their network environment by "meeting" their neighboring routers. This is done through a hello packet, which tells the neighboring router what networks the first router can reach. When the introduction is complete, the neighboring router sends the new network information to each of its neighboring routers by using a link-state advertisement (LSA). The neighboring routers copy the contents of the packet and forward the LSA to each attached network except for the one on which the LSA was received. This process is known as flooding. A router that uses a link-state routing protocol builds a tree, or map, of shortest paths, with itself as the root. The tree is based on all the LSAs seen. The tree contains the route to each destination in the network. After this tree has been built, routing information is sent only when changes to the network occur instead of periodically, as with distance-vector protocols. There are a number of advantages to link-state protocols, especially when compared to the distance vector-based routing protocols. These advantages include the following: Smaller routing tables Because the router maintains a table only of link states, rather than a copy of every route on the network, it is able to maintain much smaller routing tables. Highly scalable Link-state protocols are not limited to 15 hops as distance vector-based protocols are, so they are able to scale to much larger networks. More efficient use of network bandwidth Because link-state information is not exchanged after the network has converged, routing updates do not consume precious bandwidth, unless there is an outage that forces the network to reconverge. Faster convergence Link-state routing protocols converge faster than distance-vector protocols because updates are sent as soon as a change to the network occurs, instead of having to wait for the periodic update used with the distance-vector protocols.
One disadvantage of link-state protocols is that they are more complex to understand and configure than distance-vector protocols. They also require additional processing power on the router due to the need to calculate the routing tree. |
