What is Split Horizon?
Learn what split horizon is, how it stops distance-vector routing loops, and the poison-reverse variant — with interview Q&A.
Expected Interview Answer
Split horizon is a routing-loop-prevention rule used by distance-vector protocols that says a router must never advertise a route back out the same interface through which it originally learned that route.
In a distance-vector protocol like RIP, each router shares its known routes with its neighbors. Without split horizon, a router could learn a route from Neighbor A and then advertise that same route back to Neighbor A — and if Neighbor A later loses the real path, it might mistakenly believe this looped-back advertisement is a fresh, valid alternative, creating a two-node routing loop. Split horizon prevents this by simply suppressing that specific advertisement on that specific interface. A stricter variant, split horizon with poison reverse, still advertises the route back but marks it as unreachable (infinite metric), which converges faster because the neighbor is told immediately rather than having to wait for the route to time out. Split horizon is largely a non-issue for link-state protocols like OSPF, since every router computes routes from a shared, complete topology rather than from neighbor-reported distances.
- Prevents the most common two-node distance-vector routing loop
- Requires no extra protocol overhead — a simple suppression rule
- Poison-reverse variant speeds convergence after a link failure
- Explains why link-state protocols do not need this mechanism
AI Mentor Explanation
Split horizon is like a fielder who receives a relay throw from mid-off refusing to throw the ball straight back to mid-off — that information ("the ball is over here") clearly already came from that direction, so repeating it back would be pointless and could cause confusion about where the ball really is. Instead, the fielder only relays the position to fielders who have not already heard it. This simple rule of “never echo information back to its source” is exactly what split horizon enforces on a router interface. It stops a small, obvious redundancy from spiraling into fielders chasing a ball that has already been accounted for.
Step-by-Step Explanation
Step 1
Route learned
A router receives a route advertisement for a destination on a specific interface from a neighbor.
Step 2
Suppress the echo
The router marks that route as “do not re-advertise” on the same interface it was learned from.
Step 3
Advertise elsewhere
The router still advertises the route normally on all its other interfaces to other neighbors.
Step 4
Optional poison reverse
A stricter variant advertises the route back with an infinite metric instead of suppressing it, speeding up convergence after a failure.
What Interviewer Expects
- Correct definition: never re-advertise a route out the interface it was learned on
- Explains the two-node loop scenario split horizon prevents
- Knows the split horizon with poison reverse variant and its benefit
- Understands why link-state protocols like OSPF do not rely on this rule
Common Mistakes
- Confusing split horizon with subnetting or CIDR
- Thinking split horizon prevents all routing loops (it only stops the two-node case)
- Not knowing the poison-reverse variant exists and why it converges faster
- Assuming split horizon applies to link-state protocols the same way it does to distance-vector
Best Answer (HR Friendly)
“Split horizon is a simple rule that stops a router from telling a neighbor about a route using information that neighbor originally gave it. It is like not telling someone their own news back to them as if it were new — doing that could confuse them into thinking there is a second, independent source for something that only ever came from one place. This small rule prevents a very common and confusing type of routing loop between two directly connected routers.”
Code Example
# View split horizon status on an interface (enabled by default for RIP)
show ip interface GigabitEthernet0/1 | include horizon
# Split horizon is enabled
# Disable split horizon on an interface (rarely needed, e.g. hub-and-spoke Frame Relay)
interface GigabitEthernet0/1
no ip split-horizon
# Enable split horizon with poison reverse under RIP
router rip
version 2
no auto-summaryFollow-up Questions
- What is the difference between split horizon and split horizon with poison reverse?
- Why might split horizon be disabled on a hub-and-spoke topology?
- How do hold-down timers complement split horizon?
- Why do link-state protocols not need split horizon?
MCQ Practice
1. What does the split horizon rule prevent a router from doing?
Split horizon suppresses re-advertising a route back to the neighbor it was originally learned from.
2. What does split horizon with poison reverse do differently from plain split horizon?
Poison reverse explicitly advertises the route as unreachable rather than simply staying silent, speeding convergence.
3. Which type of routing protocol relies most heavily on split horizon?
Distance-vector protocols like RIP depend on split horizon to prevent common two-node loops.
Flash Cards
What is split horizon? — A rule preventing a router from advertising a route back out the interface it was learned on.
What loop does it primarily prevent? — The two-node routing loop between directly connected distance-vector routers.
What is poison reverse? — A variant that advertises the route back with an infinite metric instead of suppressing it, for faster convergence.
Do link-state protocols need split horizon? — No — they compute routes from a shared full topology, not neighbor-reported distances.