Fixed Wireless and LEO Satellite as ISP Backup: When to Use Them Over Broadband

A second broadband line is the default answer to ISP backup, and at many sites it is the right one. It is cheap, fast to install, and familiar to every carrier and every network engineer. The problem is that the second line usually arrives on the same street, through the same conduit, and into the same demarcation room as the first, which means the failure you are buying protection against is exactly the failure that takes both circuits down at once. When the primary and the backup share the last mile, the redundancy is accounting, not architecture, and the first backhoe, building power event, or regional fiber cut proves it.

Fixed wireless and low earth orbit satellite give you a different kind of backup because they leave the ground. A licensed point to point or point to multipoint radio reaches a tower instead of a curb, and a LEO terminal reaches a constellation instead of a central office, so both avoid the shared trench that defeats most wireline pairs. They bring their own tradeoffs in throughput, latency, weather sensitivity, and cost, and those tradeoffs determine whether either one is a smarter second path than another wire. This guide compares the three backup tiers honestly and shows when to choose a radio, a dish, or another wire.

At a Glance: Backup Connectivity Options

Why Your Second Wireline Is Often Not Diverse

Two wireline circuits are diverse only when they fail independently, and at most commercial buildings that independence is missing where it matters most. Providers lease last mile fiber from the same wholesalers, share ducts along the same street, and enter through the same penetration, so two different carrier names can ride the same physical cable for the final half mile. A single conduit cut, a single demarcation room flood, or a single provider hotel outage removes both paths even though the paperwork says you have two.

The way to catch this is to ask for diversity you can verify, not diversity you are told about. Request a route map that shows the street level fiber path, note who owns the loop from building to exchange, walk the entry points, and record which conduit and which room each circuit uses. When both feeds share any of those, the honest fix is a backup that does not use that trench at all, which is exactly where a wireless tier earns its place. A broadband second line remains sensible only when you can prove the two wireline paths actually separate before the building.

Fixed Wireless: What It Solves and Where It Struggles

Fixed wireless is a terrestrial radio link from an antenna on your building to a provider point of presence on a nearby tower or rooftop, usually one to ten miles away. Licensed bands such as 11 GHz, 18 GHz, and 24 GHz give predictable performance with coordinated spectrum, while lightly licensed 60 GHz and CBRS options offer higher capacity at shorter range with more coordination effort. Typical provisioned rates run 100 to 500 Mbps symmetric on a committed link, with some providers offering 1 Gbps under clean conditions, and one way latency to the PoP is often 5 to 15 ms, which keeps voice and interactive applications comfortable.

What makes fixed wireless valuable as a backup is not raw speed but independence. It does not use the curb, the conduit, or the building riser that your primary fiber uses, so a construction cut that isolates the street leaves the radio path intact, and a provider core outage on the wireline side does not propagate to the radio provider if you choose a different upstream. The constraints are physical. You need clear line of sight, a viable mounting point with power and grounding, and a PoP that is itself on diverse fiber and power. Heavy rain fades higher bands, foliage growth can encroach on a marginal path, and dense urban canyons can block the Fresnel zone that the link budget assumed. Where line of sight is clean and the PoP is well sited, fixed wireless is the closest thing to a diverse wire without laying new fiber.

LEO Satellite: What Changed and What Still Limits It

LEO constellations such as Starlink and OneWeb changed satellite from a last resort into a credible backup tier by moving the relay from geostationary orbit at 35,786 km to 500 to 1,200 km. The physics follows the distance. GEO round trips sit at 500 to 650 ms, which breaks interactive applications no matter how much bandwidth is available. LEO round trips commonly measure 25 to 60 ms with 50 to 250 Mbps down and 10 to 40 Mbps up on standard business plans, with flat high performance and enterprise tiers pushing higher under clear conditions. For a branch that needs to keep point of sale, voice, and remote access alive during a multi day fiber restoration, that profile is workable where GEO never was.

The limits are weather, sky view, and contention. Heavy rain and wet snow attenuate the Ka and Ku bands that LEO uses, satellite handoffs can introduce brief jitter as the terminal switches beams, and throughput varies with cell loading because the constellation is a shared resource. You need a mounting location with a wide sky view, typically 100 degrees or more, with no parapet or adjacent building blocking the scan, and you need to plan for snow, ice, and wind load on the dish. Gateway diversity matters too: your traffic lands at a ground station that connects to terrestrial fiber, so a gateway tied to the same fiber ring as your primary site reintroduces the shared fate you were trying to escape. LEO is strongest precisely where terrestrial diversity is weakest, at remote sites, temporary facilities, and any building where the street trench cannot be avoided, because it trades those local dependencies for sky view and a different ground path.

The Four Numbers That Decide the Backup Choice

Every backup decision reduces to four measurable properties, and the honest answer for any site is the option whose worst number your applications can still tolerate. First is diversity distance, how many physical and provider failure domains the backup avoids relative to the primary. Fixed wireless and LEO both clear this bar by leaving the last mile, while a second broadband line clears it only when the physical path is proven separate.

Second is latency and jitter under degraded conditions, not under the install day test. Fixed wireless holds 5 to 15 ms with little variation unless the path degrades, broadband holds 10 to 30 ms but can spike under congestion, and LEO holds 25 to 60 ms with brief excursions during handoffs or heavy rain. If your site runs real time voice, video, or transaction workloads with tight timeouts, that jitter budget is the deciding factor, and fixed wireless usually wins it.

Third is sustained throughput asymmetry. A backup sized for what must survive, not for normal peak, still needs an honest upload number: business broadband is typically 300 down and 20 up, fixed wireless is often symmetric at the provisioned rate, and LEO is asymmetric but far less so than cable. Inventory the critical applications, sum their required up and down rates, and compare that needed number to the provisioned upload of each option with headroom. Fourth is restoration independence, whether the backup stays up when the event that took the primary down is still ongoing. LEO keeps the widest independence from local infrastructure, fixed wireless keeps strong independence if the PoP is diversely fed, and a second wireline keeps only the independence the trench actually provides.

Sizing, Failover, and Operations

A backup circuit is sized for what must survive the outage, not for what the primary carries on a normal afternoon. List the traffic that cannot stop: voice, point of sale, building systems, monitoring, and remote access, estimate the aggregate up and down rate with headroom, and provision the backup for that number. A branch with a 500 Mbps primary often needs only 30 to 80 Mbps of survivable throughput, which puts fixed wireless and LEO comfortably in range if the plan you buy actually guarantees the upload. Size again for duration: a backup that holds 30 percent of normal traffic for three days is worth more than one that carries the full load for twenty minutes before its data cap or fair use policy throttles it, so check caps, prioritization, and rain fade margins before you commit.

Failover is a routing problem, not a modem problem. A second circuit does not take over by itself. Use a probe driven default route, BGP with local preference, or an SD WAN edge that steers by application, and exercise the chosen mechanism quarterly by pulling the primary during a maintenance window and measuring how long the shift takes and what degrades. The same test should verify that monitoring, remote access, and carrier escalation still work when the primary is gone, because those are exactly the tools you need during the event the backup exists to survive. Record the results, including the observed failover time and any capacity or routing fix, so the next change to the site does not silently break the pairing.

Operations also decide whether the backup is reachable when you need it. Keep modem, radio, or terminal credentials, mount location, power source, and provider contacts in the same inventory that holds the primary circuit, not in the tools that depend on the connection that just failed. Track monthly utilization on the backup path even when it carries no production traffic, because a backup that has sat idle for six months can fail quietly and the only evidence is the flat line on the graph.

Making the Choice for Your Site Mix

For most portfolios the answer is not one technology but a mix driven by site constraints. If the building offers clean line of sight and a provider can deliver a licensed link to a diversely fed PoP, fixed wireless is the strongest backup for latency sensitive and throughput hungry sites: offices, clinics, manufacturing, and any location where voice quality during failover matters. If the site is remote, newly built, in a market with poor fiber diversity, or housed in a building where no credible second trench exists, LEO is the pragmatic choice, because sky view is easier to secure than a new duct.

Where a second broadband line is genuinely diverse, meaning you have walked the entry points and verified that the provider, loop owner, and physical route separate before the building, broadband remains the cheapest and highest throughput backup and is the right call on cost alone. Where that verification fails, spending more on a radio or a dish is not overspending, it is buying the independence that the second wire cannot provide. Many teams carry all three across their footprint: broadband where the trench is proven diverse, fixed wireless where the skyline allows it, and LEO everywhere the ground does not cooperate.

Documenting the Backup You Actually Deploy

A redundancy design is only as durable as its record. The moment a backup circuit is installed, its relationship to the primary should be explicit: which site it protects, which failure modes it covers, what capacity it is expected to carry, and in what order failover occurs. Physical details matter as much as logical ones: entry point, mount location, sky view or azimuth, PoP identifier, gateway, and the power feed that keeps the backup alive when the building feed trips. Without that, a future port move or hardware refresh silently breaks the diversity assumption and the next outage reveals a backup that looks correct in the diagram but fails with the primary.

Keeping that record alongside the rest of the infrastructure is what prevents drift. Circuit groups that separate primary, broadband, fixed wireless, and LEO tiers make the design legible without a separate map, and a change history that shows who moved which circuit and when keeps the accountability visible. A site inventory that links circuits to the sites and devices they serve lets an engineer answer, during the outage, which path is active, which backup should have taken over, and whom to call when it did not. That is the practical payoff of documenting wireless backups with the same rigor as wireline ones: the decision you made about when to use a radio or a dish over another wire stays true long after the install crew leaves.