LEO can be a great safety net, but it is not a drop-in replacement for dual-entrance fiber. Treat it as a tertiary path that keeps card authorizations, tickets, and VPN sessions alive during a cut. The guide below gives you eight practical field tests, ready to run at any site, with clear pass criteria and quick calculators so you can accept the installation with evidence.
Quick position
Use case
LEO = tertiary. It keeps core apps alive during fiber cuts or construction, not as a primary or as a substitute for dual entrances.
Targets
Failover under 1 s to stable LEO; jitter p95 under 40 ms; loss under 0.5%; VPN throughput that meets your minimum app set.
Treat wireless as grace. The workhorse is still dual-entrance fiber with real route diversity.
Test matrix at a glance
| Field test | How you run it | Pass criteria |
|---|---|---|
| Failover timer | Pull primary entrance during live traffic while monitoring | Cutover to LEO ≤ 1 s with no app errors |
| Loss & jitter under load | Run upload while measuring probes | Loss ≤ 0.5% ; jitter p95 ≤ 40 ms |
| Encrypted throughput | IPsec/SSL tunnel speed test | ≥ required Mbps for apps at 70% peak |
| CGNAT & ports | Test outbound and callback flows | All target apps work; no blocked ports |
| Cloud path delta | Compare SaaS RTT via fiber vs LEO | RTT delta ≤ 80 ms for critical apps |
| Obstruction/weather | 360° sky check + snow/ice plan | Clear arcs; mount & heater plan in place |
| Power runtime | Battery sizing for CPE + LEO | ≥ 4 hours of backup power |
| Stability over time | 14-day hourly trend logs | No recurring congestion failures |
1️⃣ Failover timer: does it switch fast enough
Failover must be measured in milliseconds, not minutes. Tune detection windows and routing timers so a cashier never notices.
Failover planner
Predicted cutover
—
Result
—
Hint
—
Run a real transaction or critical app during the pull test and record screenshots.
2️⃣ Loss and jitter under load
LEO can vary with contention and beam conditions. Measure during uploads so you see the real experience.
How to run
- Start a constant upload at 60–70% of LEO’s rated uplink
- Probe loss and jitter every second to at least two targets
- Repeat during your actual busy hour
Targets
- Packet loss ≤ 0.5% average
- Jitter 95th percentile ≤ 40 ms
- No multi-second gaps in probes
3️⃣ Encrypted throughput in the real world
Spec sheets show raw speed. Your apps ride inside tunnels with overhead and CPU limits.
Encrypted throughput reality check
Estimated usable
—
Pass
—
Tip
—
4️⃣ CGNAT and ports: do your apps still talk
Most LEO consumer and business plans use CGNAT. Many outbound apps work, but callbacks, inbound checks, or pinned IP rules can break.
Checklist
- Outbound HTTPS and DNS
- VPN types you use (IPsec/IKEv2, SSL/TLS)
- VoIP or video callbacks
- Any app with IP allowlists
Application readiness score
Score / 100
—
Assessment
—
Hint
—
5️⃣ Cloud path delta: how much slower are SaaS calls
Some SaaS regions are far from your LEO gateway. Measure the RTT difference and decide which apps you allow on LEO.
Cloud RTT delta helper
RTT delta
—
Policy
—
Hint
—
6️⃣ Obstruction and weather tolerance
Terminals need a clear sky view. Snow, ice, or tree growth can hurt link quality right when you need it.
Site risk score
Risk 0–100
—
Assessment
—
Hint
—
Cold-weather kits, heaters, or covers may be needed. Plan maintenance for seasonal foliage growth.
7️⃣ Power runtime: does backup outlast the outage
A great wireless path is useless if the terminal and CPE lose power. Size battery properly.
Runtime calculator
Estimated runtime
—
Pass threshold
—
Hint
—
8️⃣ Stability over time
One clean demo is not proof. Trend lines over two weeks show whether congestion windows or thermal throttling appear at certain hours.
What to log
- Hourly RTT, jitter p95, and loss to two targets
- Upload and download at 10-minute windows
- Any session resets in VPN or SD-WAN controller
Acceptance evidence you should keep
- Failover screenshots showing cutover within target milliseconds
- Four screenshots of loss and jitter under upload load during busy hour
- Encrypted throughput test results with device CPU utilization
- Photos of mount, sky view, heater or snow plan, and cable run
- Battery runtime test or UPS meter capture
- 14-day trend graphs of RTT, jitter p95, loss, and tunnel resets
Run these tests at one pilot site before you scale LEO across the fleet. Keep the evidence pack with screenshots, logs, and photos in the as-built, and set clear steering rules so only the right apps move to LEO during a cut. When the numbers look good, LEO is a smart third path; when they do not, fix the fiber design first.
