Quick Answer
Start with the normal-state formula:
For equal-rate uplinks, normal capacity is Cnormal = N × R; after one member fails, CN-1 = (N − 1) × R. Calculate each line-rate ratio as access line rate ÷ uplink capacity, and calculate headroom as surviving capacity − busy-hour demand.
48 × 1G access = 48 Gbps
Normal uplink: 2 × 10G = 20 Gbps → 2.4:1
N-1 uplink: 1 × 10G = 10 Gbps → 4.8:1
Use the NetEngineerLab calculator with your actual port count and uplink design.
Open the Calculator →What Does N-1 Mean?
N-1 means the network is evaluated with one specified failure event. State the failure domain before counting capacity: one cable/optic/member, one line card, one upstream switch, or one complete path are different scenarios. A single member failure removes that member rate; a device failure removes every link and forwarding resource attached to that device.
The purpose is not to assume that every failure must preserve full performance. The purpose is to understand the surviving capacity and decide whether that result meets the business requirement.
Normal vs. N-1 Oversubscription
| State | Downlink Capacity | Uplink Capacity | Ratio |
|---|---|---|---|
| Normal | 48 Gbps | 20 Gbps | 2.4:1 |
| N-1 | 48 Gbps | 10 Gbps | 4.8:1 |
Use the same access-side numerator in both states; only count links that can forward traffic in the modeled state. The line-rate ratio is not a traffic forecast and does not promise per-flow throughput.
Three Worked Busy-Hour Examples
For each example, demand is an explicit planning input. Headroom equals surviving capacity minus demand; a negative result is a modeled shortfall. These examples are arithmetic illustrations, not industry utilization standards.
Example 1: 2×10G serving 48×1G access
Theoretical ratio is only the first step. Compare the surviving uplink with expected busy-hour traffic.
Planning assumption: 40 of 48 access ports are active at the modeled busy hour; 40 × 1 Gbps × 30% utilization = 12 Gbps modeled demand.
Normal capacity = 20 Gbps → headroom = 8 Gbps
N-1 capacity = 10 Gbps → headroom = -2 Gbps
Result: normal headroom is 8 Gbps; after one member fails, the 10 Gbps survivor is 2 Gbps below modeled demand. This design does not meet a no-congestion 12 Gbps failure-state objective.
Example 2: 2×25G with a 20 Gbps busy-hour requirement
Normal capacity = 2 × 25 = 50 Gbps; ratio = 48 ÷ 50 = 0.96:1.
One member fails: 1 × 25 = 25 Gbps; N-1 ratio = 48 ÷ 25 = 1.92:1.
Failure-state headroom = 25 − 20 = 5 Gbps.
This meets the stated 20 Gbps model before any separately selected engineering reserve, burst allowance or equipment bottleneck.
Example 3: 4×10G with growth and a 25% capacity reserve
Measured planning demand = 20 Gbps; forecast growth = 20%; future N-1 demand D = 20 × 1.20 = 24 Gbps.
After one member failure, surviving nominal capacity = 3 × 10 = 30 Gbps; N-1 ratio for 48×1G access = 48 ÷ 30 = 1.6:1.
For capacity reserve r = 0.25, require D / C ≤ 1 − r = 0.75, where C is the surviving usable capacity. Required usable capacity C ≥ D / 0.75 = 24 / 0.75 = 32 Gbps.
Even if usable capacity reaches the 30 Gbps nominal ceiling, utilization is 24 / 30 = 80%, leaving only 20% capacity reserve; capacity minus demand is 6 Gbps, but the shortfall against the 32 Gbps design requirement is 2 Gbps.
The 20% growth and 25% capacity reserve are scenario inputs, not prescribed standards. This design fails the selected 25% reserve requirement even at nominal capacity. Verify actual usable capacity on the surviving platform; lower effective throughput cannot improve this result. Capacity reserve is a fraction of capacity, not a percentage added to demand.
What Makes an N-1 Result Acceptable?
There is no universal ratio. An acceptable failure state depends on:
- how much busy-hour traffic must continue during the failure;
- whether temporary congestion is acceptable;
- latency and packet-loss sensitivity;
- how quickly the failed path can be restored;
- how much traffic growth is expected;
- whether the network is designed for graceful degradation or full service continuity.
N-1 with LAG
If all members forward and carry equal nominal rates, surviving aggregate line rate is the sum of surviving members. A two-member 10 Gbps LAG therefore has 10 Gbps after one member failure; a four-member 10 Gbps LAG has 30 Gbps after one member failure. This assumes the peer device, control plane, remaining members and upstream path remain available.
LAG capacity is shared across flows according to the platform hash. One flow is generally limited to one member rate, and a new hash distribution may leave some members busier than others. Validate hashing, convergence, buffers and measured utilization on the actual devices; do not treat aggregate line rate as guaranteed application throughput.
| Design | Normal Capacity | N-1 Capacity | 48×1G N-1 Ratio |
|---|---|---|---|
| 2×10G | 20 Gbps | 10 Gbps | 4.8:1 |
| 4×10G | 40 Gbps | 30 Gbps | 1.6:1 |
| 2×25G | 50 Gbps | 25 Gbps | 1.92:1 |
Busy-Hour Demand and Failure Risk
Suppose current busy-hour demand is 10 Gbps and traffic is expected to grow by 30%.
Future demand = 10 × 1.30 = 13 Gbps
2×10G normal capacity = 20 Gbps → 7 Gbps headroom
2×10G N-1 capacity = 10 Gbps → 3 Gbps deficit
A design can therefore pass today's normal-state check and still fail future N-1 requirements.
Compare N-1 Upgrade Options
| Design | Normal | N-1 | Use Case |
|---|---|---|---|
| 2×10G | 20 Gbps | 10 Gbps | Moderate access demand |
| 4×10G | 40 Gbps | 30 Gbps | More granular redundancy |
| 2×25G | 50 Gbps | 25 Gbps | Higher-capacity dual-uplink design |
Engineering Design Recommendations
- Define one failure event at a time and map every link, device and shared path it can remove.
- Check whether links share a device, power source, cable route, transport path or other critical failure domain; one event can interrupt several links, so link count alone does not prove independent redundancy.
- Use measured busy-hour percentiles and application service objectives; document burst and growth assumptions separately.
- After the capacity calculation passes, validate surviving-path effective throughput, latency, packet loss, traffic distribution and failover convergence against the required service objectives.
- Check headroom after the modeled failure, then test correlated traffic, per-member load and device forwarding limits.
- If demand exceeds survivor capacity, expect queueing, drops or reduced throughput until failover or repair; decide whether graceful degradation meets the service objective.
- Increase link rate or member count only after checking device diversity, power, optics, LAG support and the next device-level failure domain.
N-1 Capacity Planning Workflow
- Calculate installed downlink capacity.
- Measure active busy-hour traffic.
- Calculate normal-state uplink capacity.
- Remove one required failure component.
- Verify whether that event can also remove links through a shared-risk device, power source, cable route or transport path.
- Calculate surviving uplink capacity.
- Recalculate oversubscription.
- Compare surviving capacity with busy-hour demand.
- Validate throughput, latency, packet loss, traffic distribution and convergence against the failure-state service objective.
- Add future growth.
- Decide whether congestion during failure is acceptable.
- Compare upgrade options if the failure state is insufficient.
Frequently Asked Questions
Why can a network pass normal-state checks but fail N-1?
Because the same traffic demand must be carried by less remaining capacity after a failure.
Does N-1 always mean one physical link?
No. It means one required component is unavailable. The exact component depends on the architecture and resilience requirement.
Should I design N-1 for 100% of theoretical access bandwidth?
Not necessarily. Use the service objective and realistic busy-hour demand, but keep the maximum theoretical ratio visible for context.
Is 4.8:1 N-1 automatically unacceptable?
No. The ratio is only one indicator. Compare it with actual demand, application sensitivity and the required service level during failure.