Skip to content
Technologynetworking

Bandwidth Requirement Calculator

Link sizing from user count and per-user demand, separating concurrency from peak. Sizing to the average is the standard mistake: a link that carries the average comfortably is saturated at the busiest hour, which is the only hour anybody notices.

Also called: internet bandwidth calculator, link sizing calculator.

Mbps
%
%
Bandwidth required
281.3

281.3 Mbps for 50 users, of whom 30 are active at once. The average load is 150 Mbps and the peak is 225. Sizing to the average would leave the link saturated at the busiest hour, which is the only hour anyone notices. If every one of the 50 users were active at once they would have 5.63 Mbps each against the 5 assumed. Queueing delay appears well before a link is full, which is what the 25% headroom is for.

Users active at once
30
Average load
150
Peak load
225
Bandwidth each if all users are active
5.63
Data a month
16.2
On the peak
The average load is 150 Mbps and the peak is 225. Sizing to the average would leave the link saturated at the busiest hour, which is the only hour anyone notices.
On headroom
If every one of the 50 users were active at once they would have 5.63 Mbps each against the 5 assumed. Queueing delay appears well before a link is full, which is what the 25% headroom is for.
How this is calculated

What each assumption costs

StepMbpsWhat it covers
All users at full rate250The figure nobody should size to
Concurrency at 60%150Only some users are active at any moment
Peak factor 1.5x225The burst inside the busy period
Headroom 25%281.3Growth, and the delay that appears before saturation
Method and background

How this is calculated

Three multipliers sit between a user count and a link size and they are usually collapsed into one guess. Concurrency is the share of users active at any moment, which is well below one for anything interactive. The peak to average ratio then covers the burst within the busy period, since demand is not flat even among active users. Headroom covers growth and the fact that a link running at ninety percent utilisation shows queueing delay long before it shows packet loss. The page also reports what each user would get if every one of them were active at once, which is the number that decides whether the link fails gracefully or catastrophically.

users, scaled by how many are active at once, by what each one uses, by the peak ratio, plus headroom
n
Users
c
Concurrency
p
Peak to average ratio
h
Headroom

Worked examples

Each of these is asserted on every build. If a change to the engine ever moved one of these answers, the build would fail before the page could print it.

fifty users at 5 Mbps

Concurrent users
50
Bandwidth per user
5 Mbps
Share active at the same time
60%
Peak to average ratio
1.5
Headroom to leave
25%
Busy hours a day
8

Bandwidth required281.3

50 x 0.6 x 5, then 1.5x and 25% headroom

Open this example

full concurrency with no peak allowance

Concurrent users
50
Bandwidth per user
5 Mbps
Share active at the same time
100%
Peak to average ratio
1
Headroom to leave
0%
Busy hours a day
8

Bandwidth required250

boundary: with every multiplier neutral the answer is the naive product

Open this example

Method and limits

What it assumes

  • Per-user bandwidth is a sustained figure rather than a burst.

What it deliberately does not model

  • Latency and jitter matter more than raw bandwidth for interactive and real-time traffic.
  • A shared uplink is only as good as the contention ratio the provider applies upstream.

Formula version 1.0.0 · definition 1.0.0 · United States · Report a problem with this calculator

Frequently asked questions

Why size above the average?
Because nobody experiences the average. A link that carries the mean load comfortably is saturated during the busy hour, and the busy hour is when the complaints come.
How much headroom is enough?
Twenty to thirty percent is a common working figure. Queueing delay becomes noticeable well before a link reaches full utilisation, so running at ninety percent already feels slow.