Split Horizon is a technique to prevent the formation of loops while sharing the route information of the Split Horizon router. According to this rule, the Router takes a new route it has learned and first adds it to its Route Table. It transmits the current route information to other routing devices to which it is connected via a message containing the route update information. Under normal circumstances, this process is beneficial, but when routing logic is not computed, serious bottlenecks can occur in the network, causing routing loops. For this reason, it should not perform the new route created in the reverse direction over the interface that it has heard/learned. In other words, a router should not return the same information to itself over the interface to which it transmits the new route information. It does not need to receive this information message, since it already owns the change on the interface that made the route change. If the new route information comes back to the router over the same interface, a continuous routing loop occurs in the network traffic until the TTL transmission value of the packet is reset.
Especially in networks operating in the Hub & Spoke structure, that is, in the frame-relay structure, this situation causes the correct route not to be learned by all devices. For this reason, if there is a possibility that more than one network may be connected to an interface, a split horizon should be disabled. Thus, an update message coming over the same interface can be transferred to a device in another network over the same interface.
Why is Split Horizon needed? Because Distance-Vector Routing protocols such as RIP, IGRP, and EIGRP are susceptible to routing loops that occur when a data packet is caught in an infinite loop and is constantly routed through the same routers. Split Horizon emerged to avoid these loops. To give an example for RIP, since the hop count is 15, this transmission can provide a maximum of 15 loop transmissions for a packet.
Split horizon alone is not sufficient to solve this problem in all topologies. For this reason, other types of protocols, such as OSPF, use different mechanisms to prevent packet loops from occurring. In the Route Reflection technique working on BGP, while the iBGP peers transmit the new route information to each other, the updated route information is transmitted to the iBGP peers other than the router from which the information is received. This prevents loop formation. In addition, only the best routing routes are published with Route-Reflection.
Although Route Poisoning is different from its working structure, it has a similar usage purpose to Split Horizon. Route Poisoning is a method that prevents the transmission of data packets to the route that has become invalid in the network. When dynamic routing protocols detect that a route is no longer valid, it informs routers on the network that this route has an infinite metric value. A route that exceeds the maximum allowed hop count limit is of course considered unreachable. Thus, the infinite value is considered the highest and this route becomes ineffective and packet sending is blocked. This is where the poisoning of the route takes its name. We mentioned that the hop count limit for RIP is 15, if we will show this situation as an example, setting the hop count number to 16 instead of setting it to infinite on the Routing Table means that this route is not suitable for use.
