Question

Discuss routing issues and solutions namely, count-to-infinity, split horizon, split horizon with poison reverse, and hold-down timers. How have those solutions led to a better way to route information?

Answer 1

Answered accordingly. Please comment for further help.

Routing Loops is the main problem with routing, as Bellman-Ford Algorithm cannot avoid loops. This network routing loop creates the problem of Count to Infinity. Routing loops generally happen when any interface is down or when two routers simultaneously send updates.

Split Horizon

Split Horizon laws dictate that an interface from which routing data has been learned should never be sent. In Figure, R3 provides updates to R1 and R4 around IP5. R1 and R4 should announce this data to R2 and back to R3 up to this stage. This advertising method is called the "inverse route" to the originator.

Common sense dictates that R1 and R4 should not advertise back to R3 because they already know about IP5 and waste funds off-course. Split horizon protects routing loops from evading this inverse path phenomenon. For instance, let's suppose that there is no divided horizon in impact and that IP5 network is unattainable. Even if the timer for route invalidation is in effect, R3 will have to wait for the next regular update. But the R4 update comes early, unexpectedly. Now R3 can be reached in two hops via R4 in IP5 and a routing loop has taken place.

Split Horizon with Poison Reverse (route poisoning)

Uses the value of "infinity" and announces the path back to the originator on the interface from which it was taught as "unattainable." The analogy is straightforward: "There is better data than no data." Because each neighbor is publicity paths leaned on the same interface as unattainable, the count to infinity issue can be prevented. There's no need to wait to achieve infinity for hop count.

Count to Infinity

Split horizon's main purpose is to prevent neighborhood loops. But split horizon alone in a network like Figure cannot prevent loops. We don't consider various paths to the target for simplicity.

We now have the following routing table entries for IP5 if the network is stable:

Router	Network	Via	Hop Count
R3	IP5	Directly connected	0
R4	IP5	R3	1
R2	IP5	R4 (and R1)	2
R1	IP5	R3	1

Let's take a closer look at the propagation of routing data. The IP5 network failed. R3 sends an update that IP5 is unattainable to R1 and R4. The following route requests are now available for R3, R4 and R1.

Router	Network	Via	Hop Count
R3	IP5	Unreachable	Null
R4	IP5	Unreachable	Null
R1	IP5	Unreachable	Null

Since R2 has not yet notified the update that IP5 is not (and sadly) accessible now, it sends an update to R1. How has the number of hops risen? In fact, R2 learned from R1 and R4 IP5; therefore, data obtained from R4 is a distinct entity from R1. R1 advertises IP5 through R4 to R1 and vice versa, in this condition, the split horizon rule is not void. R1 and R4 will not install these ads from R2 in a stable network, as they already have the best route in one hop through R3.

Router	Network	Via	Hop Count
R2	IP5	R4 and R1	2
R4	IP5	R2	3
R1	IP5	R2	3

R1 will send the update to R3. R3 can now reach IP5 in four hops through R1. The table for the routing will be:

Router	Network	Via	Hop Count
R3	IP5	R1	4

R3 installs and advertises this path to R4. R4 now thinks that the R3 measurement has risen towards IP5 ; however, it is the only path to be installed. R4's routing table looks like:

Router	Network	Via	Hop Count
R4	IP5	R3	5

R4 advertises and goes on to R2, R2 to R1 to R3. The hop count continues to rise for each update. It is called to count this phenomenon to infinity. Defines "infinity" to elevate this vector range issue. For instance: RIP describes infinity as hop count reaches 16 and instantly discards such routing data. Although the routing loop will be removed, the convergence will be slow as each router will have to wait until the hop count reaches infinity.

Hold Down Timer

Flash (or activated) updates made quick convergence possible, but holding down timers creates uncertainty to decrease the recognition of poor routing information. If an advertisement is obtained with enhanced measure (or hop count), before accepting the fresh routing information, the router will set a hold down timer. Router will not acknowledge any path with equal or worst metric for the designated location until the hold down timer expires. Trade off is the moment of convergence vs. containing poor data about routing. Before accepting any updates from R1 and R4, in Figure, R3 must set a hold down timer.