toolbox
Home / AI, Data & Technology / Fiber Network Latency Budget Planner

Fiber Network Latency Budget Planner

The Fiber Network Latency Budget Planner builds a simple one-way and round-trip latency estimate from fiber distance plus device, processing, queueing, and engineering allowances. It is useful for early architecture comparisons, distributed systems planning, data-center interconnect discussions, or checking whether a proposed path leaves enough room for non-propagation delays.

The planner uses an approximate propagation delay of 0.005 milliseconds per kilometer of fiber, equivalent to about 5 microseconds per kilometer. You then add per-device forwarding delay, application or processing delay, queueing allowance, and an extra margin. The result is a budget estimate rather than a packet-level prediction. Real latency varies with routing, transceivers, serialization, congestion, buffering, protocol behavior, and the actual cable route, so measured or vendor-specific values should replace generic assumptions when available.

Inputs

km
devices
ms
ms
ms
ms
Result
—
Estimated one-way latency
Fiber propagation delay—
Device forwarding delay—
Other + queueing + margin—

1. Enter one-way fiber length
Use the actual or estimated cable-path distance, not straight-line geographic distance.

2. Count intermediate devices
Include switches, routers, optical platforms, or other elements for which you are applying the same average forwarding delay.

3. Set per-device delay
Use a measured or vendor-derived value when available; otherwise treat the default only as a planning assumption.

4. Add processing and queueing
Enter non-propagation delays that are outside the repeated device term.

5. Reserve margin
Add explicit headroom for uncertainty or unmodeled components.

6. Review one-way and round-trip budget
The main result is one-way latency. The breakdown also shows the implied symmetric round trip as twice that value.

Propagation delay (ms) ≈ Fiber distance (km) × 0.005 ms/km Device delay (ms) = Device count × Delay per device One-way latency = Propagation + Device delay + Processing + Queueing + Margin Estimated RTT = 2 × One-way latency

The 0.005 ms/km propagation factor is an approximation for light traveling through optical fiber. The RTT calculation assumes a symmetric return path with the same delay.

What the result means

The result is the modeled one-way latency budget for the path under the supplied physical distance and non-propagation assumptions.

Use measured path latency and actual routing for production validation, especially for applications with strict latency or jitter requirements.

Given: 320 km fiber, 6 devices at 0.08 ms each, 0.6 ms processing, 1.2 ms queueing, and 0.5 ms margin.

Calculation: Propagation = 320 × 0.005 = 1.6 ms. Device delay = 6 × 0.08 = 0.48 ms. One-way = 1.6 + 0.48 + 0.6 + 1.2 + 0.5 = 4.38 ms. Estimated RTT = 8.76 ms.

Result: 4.38 ms one way; 8.76 ms estimated RTT.

Interpretation: In this example, queueing and processing together contribute more delay than the intermediate devices themselves.

Why is fiber propagation about 5 microseconds per kilometer?

Light travels more slowly in glass than in vacuum. A common planning approximation is about 200,000 km/s in fiber, which corresponds to roughly 5 microseconds per kilometer one way.

Should distance be route kilometers or straight-line distance?

Use route kilometers whenever possible. Cable paths follow rights-of-way, ducts, buildings, and slack storage, so the optical path is often longer than map distance.

Does the device delay include serialization?

Only if your chosen per-device value includes it. This planner treats per-device delay as one combined input, so you should use a value consistent with the equipment and link rates being modeled.

Why can measured RTT be higher than twice the one-way estimate?

The real forward and return paths may differ, and queueing, routing, access links, server processing, or protocol behavior can add asymmetric delay. The 2× figure is only a symmetric planning approximation.

Is latency the same as jitter?

No. Latency is delay, while jitter describes variation in delay over time. This calculator budgets average or assumed delay components and does not model jitter distribution.