Backup internet only helps if it does not share the fault that took down the main line. Check five things: whether both paths share a conduit, pole or building entry; whether both depend on the same upstream carrier or fiber; whether the modem, fiber terminal (ONT), router and switches have backup power; how the router detects a dead link and how long it waits; and what happens to DNS, public IP addresses and VPN tunnels after the switch. Dual-WAN with two wired providers and cellular (LTE or 5G) backup each pass some of these checks and fail others.
Evidence label: Documentation-based guide. ClearTrade Report Editorial Team checked Cisco Meraki, Ubiquiti, TP-Link Omada and Netgate pfSense documentation, plus Federal Communications Commission, Cybersecurity and Infrastructure Security Agency (CISA) and National Institute of Standards and Technology (NIST) documents, on November 12, 2026. No outside professional review was performed. Products named below are documentation examples, not recommendations. This guide contains no affiliate links.
Three failover designs and what each protects
Dual-WAN with two wired providers. A router connects to two services, such as fiber from one company and cable from another. It protects against a problem inside one provider’s network, less so against damage where both cables enter the building.
Cellular backup. A built-in cellular modem or separate cellular gateway uses a radio link instead of a second cable. It avoids the building’s cable entry but brings data allowances, different public addressing and dependence on the tower’s power and backhaul.
Failover or load balancing. Ubiquiti’s “WAN Failover, Load Balancing and Port Remapping on UniFi Gateways” says “Failover mode is designed to prioritize one active WAN connection, while others remain on standby,” while load balancing “distributes new outbound connections across multiple WAN interfaces based on configurable traffic splits.” Either way, one router carries all traffic, so the router itself remains a single point of failure.
Shared paths: conduit, pole, building entry and carrier
CISA’s infographic “Risk to Critical Infrastructure: Telecommunications Central Offices” tells organizations to “Verify physical diversity and redundancy with your network providers for mission critical lines and services.” It also warns that loss of critical infrastructure services “could cause central office components to shut down due to power loss,” a fault nothing in your building can fix.
Two providers may share a utility pole, a conduit or the same building entry. Two circuits from one provider usually share even more. Netgate’s pfSense page “Multi-WAN Caveats and Considerations” covers a related logical conflict: when multiple circuits from one provider share the same subnet and gateway IP address, it suggests asking the provider to configure them “in different subnets with different gateways.”
Cellular backup can share a path too. A CISA National Emergency Communications Plan spotlight, “Improving Emergency Communications Resiliency through Redundancies”, describes how “In November 2020, a city contractor accidentally cut a fiber line that disrupted cell and Internet service” in Gunnison and Hinsdale counties in Colorado. Cell towers need wired backhaul, so one fiber cut can hit wired and cellular service together.
Ask each provider whether service arrives overhead or underground, where it enters and which central office or hub serves it.
Power: every box in the chain needs it
The FCC’s consumer page “Tech Transitions: Network Upgrades That May Affect Your Service” warns that phone service “may not work during power outages if it is not line-powered (not a copper line),” and that newer services including fiber, coaxial cable and wireless “usually need backup power.”
The FCC’s 2015 order “Ensuring Continuity of 911 Communications” (Federal Register, October 16, 2015) requires “providers of facilities-based, fixed residential voice services, which are not line powered” to offer “the option to purchase a backup solution capable of 8 hours of standby power,” and, within three years, “an additional solution capable of 24 hours of standby power.” In an example disclosure, the order says “this backup power will not also power services other than voice.” Do not assume a provider battery keeps a business data network running.
NIST SP 800-34 Rev. 1, “Contingency Planning Guide for Federal Information Systems”, lists “Appropriately sized uninterruptible power supplies (UPS) to provide short-term backup power to all system components” as a preventive control. For failover, that means each ONT or modem, the cellular gateway, the router and the switches, including any Power over Ethernet switch feeding phones or access points. UPS units add heat in a closet or cabinet; see our guide to small network racks: heat, cable management and service clearance.
How the router decides a link is down
A link can fail hard (no link light) or soft (link up, no traffic passing).
- Cisco Meraki MX. “Connection Monitoring for WAN Failover” describes DNS, internet (ping and HTTP) and ARP tests. If no physical link is detected, “the failover would take place immediately.” For a soft failure, “it can take approximately five minutes for failover to occur.”
- Ubiquiti UniFi. The UniFi article says the gateway “performs continuous health checks using three probes” and “A WAN is marked down if any two out of three probes fail.”
- TP-Link Omada. “How to Configure Online Detection for WAN Link Testing” says “Online Detection results will influence whether Load Balancing and Link Backup features take effect,” and “The smaller the online detection interval, the faster Load Balancing and Link Backup features will respond.”
- Netgate pfSense. “Gateway Settings” says “By default, the gateway monitoring daemon will ping the gateway IP address.” Default packet loss thresholds are “From 10 and To 20” percent, and results are averaged over “60000 (60 seconds, one minute).” The “Multi-WAN Terminology and Concepts” page explains that with a monitor IP set, “the firewall will ping this IP address and if it stops responding, the gateway is marked as down.”
A check that only watches the link light or the first hop can miss an upstream fault. Faster checks react sooner but can flap on a shaky line.
DNS, public IP addresses, port forwards and VPN
Public IP address. After failover, traffic leaves from the backup provider’s address. Port forwards, VPN peers and vendor allowlists tied to the main address need a plan. Netgate’s “Multi-WAN and NAT” page says “Each port forward applies to a single WAN interface.”
Cellular addressing. Meraki’s “Carrier-Grade NAT and Meraki Auto VPN” says “Many carriers use CG-NAT (Carrier-Grade NAT) within their networks” and “This is primarily seen on cellular deployments.” It adds that CG-NAT “can end up causing issues with Auto VPN as the IPs/Ports in use will vary.” Ask the carrier whether a static public IP is offered on your plan.
DNS. Netgate’s “Interface and DNS Configuration” says “Most ISPs prohibit recursive queries from hosts outside their network.” It advises “At least one gateway from each WAN should be selected where possible.” If every DNS server belongs to the primary provider, lookups can fail on the backup.
Sessions and failback. Expect calls and VPN sessions in progress to reconnect. Return can be gradual. Meraki describes “Graceful failback behaviour,” in which “All traffic with an existing mapping will continue to use the secondary uplink.” On a metered cellular backup, that traffic keeps counting against the plan.
Cellular data use and plan limits
Monitoring itself uses data. Meraki’s connection monitoring page says “Cellular connection testing is reduced in an effort to minimize the overall data usage on the link,” and that a cellular backup runs its DNS queries and HTTP checks “every 4 hours.” A broken backup can go unnoticed for hours. Read the plan’s data allowance and speed reduction terms. For traffic pinned to one WAN by policy, the UniFi article says “To allow failover for this traffic if the WAN goes down, disable Kill Switch in the policy configuration.”
Failover comparison table
| Failure point | Dual-WAN, two wired providers | Cellular backup | What to check | Source |
|---|---|---|---|---|
| Cable cut at shared conduit, pole or building entry | Both lines can fail if they share the path | Usually survives, unless the cut also hits tower backhaul | Ask each provider how service reaches the building | CISA central offices infographic; CISA NECP spotlight |
| Upstream fiber or central office outage | Survives only if providers use different upstream paths | Can fail if towers use the same fiber | Verify physical diversity with providers | CISA central offices infographic; CISA NECP spotlight |
| Building power outage | Fails unless ONT or modems, router and switches have backup power | Fails unless the cellular gateway, router and switches have backup power | Put every box in the chain on a UPS and time the runtime | FCC; NIST SP 800-34 |
| Soft failure (link up, no internet) | Depends on probe settings | Depends on probe settings | Probe outside targets, not only the first hop | Meraki; UniFi; Omada; pfSense |
| Inbound services and VPN | Backup address differs; needs rules per WAN | Carrier-grade NAT is common on cellular | Plan port forwards, VPN peers and allowlists for each WAN | pfSense; Meraki |
| DNS | Provider DNS may refuse queries from the other network | Same issue | Bind DNS servers to gateways or use resolvers reachable from both | pfSense |
| Data use after failback | Usually no data cap issue | Traffic can stay on cellular after the primary returns | Watch usage and failback behavior | Meraki |
| Router failure | Single point of failure | Single point of failure | Plan a spare or high availability pair | Our summary |
This table is our summary of the cited documents.
How to test the backup you have
NIST SP 800-34’s sample policy says “The plan recovery capabilities and personnel shall be tested annually to identify weaknesses of the capability,” and lists “Internal and external connectivity” among the areas tests should cover. CISA’s spotlight urges organizations to “regularly revisit, test, and update” resiliency plans “to ensure continuity of service.” A practical test, scheduled outside business hours:
- Hard failure. Unplug the primary WAN cable at the router. Time how long until browsing, phones and card terminals work.
- Soft failure. Leave the router cabled to the modem but break the provider side, for example the coax at a cable modem, so the router keeps its link light. Ask your installer before handling fiber connections.
- Power. Run on battery and note which device shuts off first.
- Services. Check remote access, VPN, allowlisted apps and DNS.
- Failback. Record when traffic returns and how much backup data was used.
Know the limits of this guide
This guide is based on vendor documentation and government documents, not on tests by ClearTrade Report. Vendor defaults change with firmware, so check documentation for your model and version. The FCC backup power rule cited covers residential voice service, not business data service. NIST SP 800-34 targets federal systems and CISA examples involve public safety networks; we cite them as general practice. This guide is not engineering advice. Your providers or installer can confirm paths and power at your site.
Primary references checked November 12, 2026: Cisco Meraki, “Connection Monitoring for WAN Failover”; Cisco Meraki, “Carrier-Grade NAT and Meraki Auto VPN”; Ubiquiti, “WAN Failover, Load Balancing and Port Remapping on UniFi Gateways”; TP-Link Omada, “How to Configure Online Detection for WAN Link Testing”; Netgate, pfSense “Gateway Settings,” “Multi-WAN Terminology and Concepts,” “Multi-WAN Caveats and Considerations,” “Multi-WAN and NAT” and “Interface and DNS Configuration”; Federal Communications Commission, “Tech Transitions: Network Upgrades That May Affect Your Service”; Federal Communications Commission, “Ensuring Continuity of 911 Communications,” Federal Register Vol. 80, No. 200 (October 16, 2015); CISA, “Risk to Critical Infrastructure: Telecommunications Central Offices”; CISA, “Improving Emergency Communications Resiliency through Redundancies” (NECP Spotlight); NIST SP 800-34 Rev. 1, “Contingency Planning Guide for Federal Information Systems” (May 2010, updated November 11, 2010).
If your phones run over the same connection, read our checklist on VoIP emergency calling, location records and power dependencies before a failover plan is treated as complete.
Frequently asked questions
Are two internet providers enough for business failover?
Not by themselves. Two providers can still share the same conduit, pole, building entry or upstream fiber, so one cut or one outage can take down both. CISA advises organizations to verify physical diversity and redundancy with their network providers for mission critical lines and services. Ask each provider how its service reaches your building.
Will my internet stay up in a power outage if my provider sold me a battery?
Not necessarily. The FCC backup power rule from 2015 covers residential voice service that is not line powered, and its example disclosure says the backup power will not power services other than voice. Your modem or ONT, router and switches need their own backup power.
How fast does WAN failover happen?
It depends on how the router detects a failure. Cisco Meraki documentation says failover is immediate when no physical link is detected, but can take approximately five minutes for a soft failure where the link stays up. UniFi, Omada and pfSense let you adjust probe targets, thresholds or intervals. Test your own setup and time it.
Why does remote access or a VPN stop working on the cellular backup?
The backup connection has a different public IP address, and inbound rules tied to the primary address will not follow it. Cisco Meraki documentation says many carriers use carrier-grade NAT, primarily on cellular, which can cause VPN issues because the IP addresses and ports in use vary. Ask the carrier about static public IP options.
Disclaimer: This guide is general information based on equipment vendor documentation and U.S. government documents, not professional engineering, IT or legal advice for a specific site. AI tools assisted with drafting. ClearTrade Report Editorial Team checked the cited sources and edited this guide.
Need a qualified professional?
Trusted Pros uses visible license-verification states and never ranks paid listings above unpaid listings.
