Subscriber Capacity & Oversubscription Calculator

Model how many subscribers an uplink really supports using peak-hour concurrency and average usage, rather than a contention ratio someone invented.

Calculator ISP Toolkit Runs in your browser
Share generating traffic in the busy hour.
Residential busy-hour averages are typically 3-12 Mbit/s.
Try:

Result

Peak-hour demand
6.40 Gbit/s
800 active subscribers
Uplink utilisation at peak
64%
Nominal contention ratio
100:1
sold capacity vs uplink
Effective headroom factor
1.6×
uplink vs real peak demand
Subscribers supportable
2,187
keeping 30% headroom
If every active subscriber pulled at once
12.5 Mbit/s
each
Peak-hour utilisation6.40 Gbit/s of 10.0 Gbit/s

Sensitivity to concurrency assumptions

ConcurrencyPeak demandUtilisationMax subscribers
20%3.20 Gbit/s32%4,375
30%4.80 Gbit/s48%2,916
40%6.40 Gbit/s64%2,187
50%8.00 Gbit/s80%1,750
60%9.60 Gbit/s96%1,458
70%11.2 Gbit/s112%1,250
Within capacity
Comfortable. Revisit the concurrency and average-demand assumptions quarterly against real interface graphs - both drift upward over time.
Why ratios mislead
Contention ratio as a headline number is close to meaningless: 2000 subscribers on 500 Mbit/s plans is nominally 100:1, yet if busy-hour demand averages 8 Mbit/s across 40% of them the uplink barely notices. Plan against measured busy-hour throughput, not against the sum of what you sold.

About Subscriber Capacity & Oversubscription Calculator

Oversubscription is unavoidable and, done with real numbers, entirely safe. The failure mode is planning against sold capacity - or against a contention ratio inherited from a decade ago - instead of against measured busy-hour demand.

Measuring instead of guessing

The two inputs that drive this model — busy-hour concurrency and average demand per active subscriber — are measurable, and measuring them beats any industry rule of thumb. Take 95th-percentile interface throughput at the aggregation point over a month, divide by the subscriber count on that interface, and you have your real per-subscriber busy-hour figure. Track it quarterly: it has risen every year for two decades, and a capacity plan built on last year’s number is already behind.

The numbers that actually matter

Two figures drive everything: what fraction of subscribers are active in the busy hour, and what each of those averages. Residential broadband typically sees 30-50% concurrency at peak with 3-12 Mbit/s average demand, dominated by adaptive-bitrate video which backs off gracefully. Business connections run higher concurrency and are far less tolerant. Plan speed barely enters the calculation - a 1 Gbit/s subscriber watching television consumes 15 Mbit/s.

Common use cases

  • Deciding when an uplink upgrade becomes necessary.
  • Sizing a new PoP or sector before committing to hardware.
  • Testing how sensitive a plan is to concurrency assumptions being wrong.

Edge cases and gotchas

  • Busy-hour demand grows year over year - build in growth, not just headroom.
  • A single large game or OS release can double peak demand for a day; headroom is what absorbs it.

Frequently asked questions

Why is the 95th percentile used rather than the peak?
It discards brief spikes that no sane design pays for while still capturing sustained demand. It is also how transit is billed, which makes it the natural planning unit.
What contention ratio should I use?
None - the ratio is an output, not an input. Measure your busy-hour throughput per subscriber from interface counters and plan against that. Ratios only make sense when comparing two networks with identical usage patterns.