Network Capacity Forecast Planner: Busy-Hour Growth and Expansion
Turn measured busy-hour utilization into a growth forecast, an expansion trigger and a capacity target.
Measured baseline and design assumptions
Use one consistent resource boundary such as a link, port-channel, path or shared pool. The loaded numbers are an engineering example, not a measured production link.
Historical busy-hour observations (optional)
Enter valid values. The expansion threshold must be above current utilization and no higher than installed capacity.
Core calculation
current demand = installed capacity × utilization; current peak = current demand × (1 + peak factor); threshold = installed capacity × target utilization; years to threshold = ln(threshold / current peak) / ln(1 + annual growth); required capacity = projected peak ÷ target utilizationForecast and expansion decision support
— —
Annual forecast
| Year | Demand | Peak | Peak utilization | Headroom | Status |
|---|
Engineering boundaries
This browser-only planner does not infer traffic from raw counters, model seasonality, queue behavior, ECMP imbalance or failure topology. Its growth projection is a screening model; validate with telemetry and a design review.
How to use the forecast
Use a measured baseline
Use busy-hour P95/P99 or another documented planning percentile from a representative period. Keep the capacity boundary and traffic direction consistent.
Choose a trigger and horizon
The threshold is an operational planning trigger. Set it from latency, loss, queueing, service commitments and procurement lead time rather than assuming a universal safe percentage.
Read the expansion date
Months to threshold assumes a constant compound growth rate. Treat it as a planning estimate and update it as measured growth changes.
Validate the capacity plan
Review surviving capacity under failures, path diversity, traffic asymmetry, hardware limits, queue drops and link procurement lead time before approving an upgrade.
Network capacity planning method
Use a consistent link, port-channel or shared-pool boundary. Current demand is C × U; peak demand is C × U × (1 + P), where C is installed capacity in Gbps, U is measured utilization and P is the peak allowance. A forecast at year t uses Dt = D0(1 + g)t. Required capacity at the target utilization T is Dt/T.
Example: 100 Gbps at 50% busy-hour utilization, 20% annual growth and 10% peak allowance gives 55 Gbps current peak; after three years it is about 95.04 Gbps. At a 70% target, plan about 135.77 Gbps of usable capacity. Recalculate with measured P95/P99 telemetry before procurement.
Limits and review questions
This screening model does not infer seasonality, queue drops, ECMP imbalance or failure topology. Check surviving capacity, path diversity, asymmetry and lead time separately. See the data-center capacity topic and bandwidth calculator.
FAQ
Is theoretical capacity the same as busy-hour load? No. Port speed is a ceiling; the measured percentile and protocol overhead determine planning demand.
When should I set the expansion threshold? Use latency, loss, queueing, service targets and upgrade lead time; there is no universal safe percentage.
Frequently asked questions
How should I choose the annual growth rate?
Fit a trend to several months or quarters of comparable busy-hour measurements and document known changes such as new sites, users or applications.
Does the expansion month guarantee there will be no congestion?
No. It is based on the supplied average baseline and growth assumption. Short peaks, imbalance or failures can exceed the threshold earlier.
Why can required capacity exceed installed capacity well before the horizon?
The plan reserves the specified target utilization as headroom. Required capacity is forecast peak demand divided by that target.
Method: compound growth and target-utilization sizing. Review against live telemetry before procurement.
Switch capacity planning topicPart of the switching and bandwidth calculators hub.
ENGINEERING WORKFLOW
Continue the engineering workflow
Plan addressing, VLAN capacity, packet size, uplinks, optics, access policy and DNS behavior as one LAN design sequence.