A fiber cut is not rare enough to ignore, and it rarely fails the way people picture it. The most painful outages are the ones that take out two “redundant” circuits at once because they share a conduit, a pole line, a building entry, or an upstream aggregation point. True resilience comes from designing around shared risk, not just buying a second bill.
Before the 16 designs: the one rule that matters
Redundancy is two ways to get online. Diversity is two ways that do not fail together. Most “fiber cut” failures happen because both paths share the same risk.
Quick glossary (plain-English)
- Shared risk: two “separate” circuits that still share a conduit, pole segment, manhole, cabinet, or upstream hub and can fail together.
- Dual entrance: two physically separate building entry points so a single cut at the wall does not end both paths.
- Active active: both links carry traffic, often with policy steering and health checks.
- Active standby: backup link stays idle (or low use) until the primary fails.
- Last mile: the final stretch from provider plant to your building. This is where many “double outages” happen.
Failure patterns these designs are built to survive
- Single backhoe event takes out a shared conduit or shared trench
- Single pole failure or storm event impacts an aerial span used by both circuits
- Single building entrance or riser incident drops both “redundant” paths
- Provider aggregation fault takes out both circuits because they converge upstream
- Brownouts: link stays up but quality collapses and apps fail
16 real-world designs that survive fiber cuts
1️⃣ Dual ISP, dual entrances, documented separation to the street
Two providers, two separate building entry points, and a written statement of how each circuit reaches the street.
The goal is not “two contracts.” The goal is two physical paths that diverge immediately.
Survives
Building entrance cuts, single conduit cuts near the wall, single handhole failures close to the site
2️⃣ Same ISP, truly diverse laterals plus different upstream hubs
Sometimes you must stay with one carrier. This design only works if you require two laterals plus diversity beyond the last mile,
including different upstream aggregation where possible.
Reality check
If both circuits converge to the same upstream hub or share the same street route, this becomes “two bills, one failure.”
3️⃣ Fiber plus fixed wireless access as a physical-damage hedge
One path is fiber. The other is a different medium entirely, often fixed wireless. The purpose is to survive a local civil event
that wipes out underground plant in one area.
Best fit
Retail, clinics, offices that can run “degraded but online” during an incident
4️⃣ Fiber plus LTE or 5G, tuned for critical apps only
This is not “backup internet for everything.” It is “backup internet for what must stay alive.”
Route only the essentials: POS, payment terminals, EHR access, support tools, core VPN, and a slim collaboration profile.
Key tactic
Predefine a failover policy set that automatically disables bandwidth-heavy traffic when on cellular.
5️⃣ Dual routers, dual power, dual WAN, active standby edge
Plenty of “fiber cut” incidents are actually edge failures. This design survives a router failure, a power supply failure,
or a misconfiguration push, while still allowing you to fail over between circuits.
Survives
Router failure, single UPS event, single WAN port failure, plus the external circuit loss
6️⃣ SD WAN with health checks, active active steering
Both links carry traffic, but the platform constantly tests quality and steers traffic away from a brownout.
This is how you avoid “the circuit is up but everyone is down.”
Make it real
Failover detection should test multiple targets so you do not fail over to a worse path because one probe went weird.
7️⃣ BGP multihoming for inbound plus outbound resilience
If you host services or need predictable inbound access, BGP multihoming can keep inbound routes alive even when one ISP fails.
This is more complex but it solves a problem that simple failover often does not: inbound reachability.
Best fit
Businesses with public services, strict VPN needs, or sites that act like small data centers
8️⃣ Dual circuits to two different cloud onramps or PoPs
Many teams buy redundancy at the building but still send both paths to the same upstream cloud entry.
This design ensures your “two ways to the cloud” are actually two ways to the cloud.
Practical outcome
Lower chance of both links being affected by a single provider peering or regional incident
9️⃣ Two providers, one is fiber, one is coax, with strict shared-risk checks
Coax can be a useful hedge if it truly follows a different path and terminates in different upstream plant.
The value is physical and operational separation, not marketing labels.
Avoid
Two services that still enter the building through the same conduit or share the same MPOE pathway
🔟 Dual last mile plus dual internal distribution, separate risers
Plenty of failures occur inside the building: a riser issue, accidental cut, or a single telecom room incident.
This design uses physically separate internal routes to keep at least one path alive.
Best fit
Multi-floor offices, hospitals, campuses, warehouses with long internal runs
1️⃣1️⃣ Branch plus local breakout, no single corporate hub dependency
If every site depends on a single corporate hub, a hub outage becomes a company outage.
This design allows local internet breakout for SaaS while preserving secure paths for what must traverse the core.
1️⃣2️⃣ Two sites, mutual backup, each can run core services
Instead of trying to make one site perfect, run two sites that can temporarily carry each other.
This design often pairs well with active active internet and a clear “degraded mode” runbook.
1️⃣3️⃣ DNS and identity resilience built into failover
Many failovers work at the packet level but fail at the user level because DNS, identity, or VPN endpoints are not reachable.
This design ensures DNS resolvers, identity providers, and VPN endpoints are reachable from both paths.
1️⃣4️⃣ Voice and video survival profile: separate policy for real-time
Real-time traffic is sensitive to loss and jitter. This design uses policy to keep voice stable during failover and brownouts,
even if it means downgrading everything else temporarily.
1️⃣5️⃣ “Two circuits, one router” replaced with “two circuits, two edges”
A single edge is a single point of failure. This design uses two independent edge devices, ideally with separate power sources,
and validates that either edge can carry the full critical load.
1️⃣6️⃣ The “shared risk audit” design: prove diversity before you celebrate
This is the design most teams skip. You do a documented shared-risk audit across building entry, street route, pole spans,
handholes, and upstream convergence. You do not label it “diverse” until you can defend it on paper.
Proof points to request
- Separate building entrances and separate internal pathways
- Separate street routes to first aggregation point
- Different upstream hubs where feasible
- Clear statement of any unavoidable shared segments
Design picker table (fast way to match a business to the right pattern)
Failover reality calculator (how much does “slow failover” cost?)
This turns your assumptions into a monthly “failover drag” number. It helps justify spending on better diversity, better detection,
and policy-based steering that avoids brownouts.
Results
Annual cost of slow failover
$0
Monthly cost of slow failover
$0
If this number is meaningful, prioritize designs that improve detection, avoid shared risk, and steer around brownouts.
The short checklist that prevents fake diversity
- Two separate building entrances and two separate internal pathways
- Written acknowledgment of any shared conduit or shared pole segments
- Different upstream hubs where feasible, not just different last mile brands
- Failover testing that includes brownout detection, not only link down
- A degraded-mode policy set so the business stays productive on backup links
If you want redundancy that actually survives fiber cuts, design for shared-risk separation first, then pick the technology and routing approach that matches your operational tolerance for degraded performance during an incident.
