Glossary
The 95/5 Method: How Datacenter Traffic Billing Really Works
What percentile billing means in practice, why spikes are ignored, and how CDR commitments keep the model fair for everyone.
Published: December 14, 2021 · Updated June 4, 2026
Deutsche VersionIf you connect to the internet through colocation, cloud uplinks or IP transit, your invoice rarely reflects “every byte exactly as it happened.” Providers use different models—and in datacenters the 95/5 percentile method has been the de-facto standard for years.
At Vortanix we bill many connectivity products the same way our upstream carriers do: transparently, on measured Mbit/s rates, with room for real-world spikes. This article explains what that means for your monthly costs—and what you should watch for in your contract.
Two ways to bill traffic
Volume-based billing charges transferred data (gigabytes or terabytes). You may have a monthly allowance; overages are billed per GB. Some products behave like a mobile flat rate with a fair-use cap.
Rate-based (bandwidth) billing looks at how full your line was over time. Instead of punishing every split-second peak, the industry uses percentile billing—and the most common variant is 95/5.
What is the 95/5 method?
The name is literal: during a billing period (usually one calendar month), traffic is measured continuously. At the end, the highest 5 % of samples are discarded. The highest value that remains is your billable Mbit/s rate for that month.
The idea is simple: networks breathe. Backups, patch Tuesdays, marketing campaigns and DDoS scrubbing can create short spikes that do not represent your normal operation. Percentile billing ignores those outliers so you are not charged as if every spike were your new baseline.
How billing works in practice
Typical measurement setup:
- Interval: one sample every five minutes (in + out are often combined or tracked separately—check your contract).
- Samples per day: 288 (12 per hour × 24).
- Samples per 30-day month: about 8,640.
- Discarded (top 5 %): roughly 432 highest samples—equivalent to ~36 hours spread across the month.
After removing those peaks, the next-highest sample defines what you pay for, multiplied by your contracted price per Mbit/s. Your port might be 1 Gbit/s, but you only pay for the utilised percentile rate—not the physical ceiling.
Tip for operators: ask whether measurements are taken on the hand-off port, after shaping, or aggregated across multiple VLANs. At Vortanix we document the measurement point in your service description so there are no surprises when you compare graphs to invoices.
Why customers prefer it
Percentile billing matches how capacity is actually planned. A hosting provider or enterprise WAN team provisions for sustained load plus headroom—not for a five-minute backup burst at 3 a.m.
Compared with pure peak billing, 95/5 typically means:
- More predictable invoices when traffic is uneven.
- Less penalty for maintenance windows and batch jobs.
- A single number (Mbit/s) that is easy to compare across months.
It is not a free lunch: if your baseline rises, the percentile rises with it. The model forgives spikes—not sustained growth.
CDR: the safety net
Could a customer sign five contracts and burst two days at each provider while idling the rest of the month? Theoretically yes—which is why serious contracts include a CDR (Committed Data Rate).
A CDR is your minimum committed utilisation (or minimum invoice basis) for the billing period. Even if your 95th percentile drops to zero, you still pay at least the CDR. Common values sit around 10 % of port speed, but every project is different.
For you as a customer, the CDR is the incentive to spread load sensibly: bursting on “free” percentile headroom still costs something if you fall below your commitment. For the provider, it covers cross-connect, port and transit costs that exist whether you burst or not.
When reviewing a Vortanix quote, compare port size, CDR, price per Mbit/s and measurement direction (inbound, outbound or sum)—not just the headline port price.
A practical example
You have a 500 Mbit/s IP-transit port. Most business hours you sit around 120–180 Mbit/s. Every Sunday night a backup pushes you to 420 Mbit/s for four hours. One Tuesday, a misconfigured sync hits 480 Mbit/s for twenty minutes.
Under 95/5, those Sunday and Tuesday spikes likely land in the discarded top 5 %. Your bill might be based on ~190 Mbit/s—not 480 Mbit/s. If your CDR is 50 Mbit/s (10 % of port), you are well above it; the percentile drives the invoice.
If, instead, you run flat at 400 Mbit/s all month, the percentile catches up with you—that is sustained demand, not a forgiven spike. Capacity planning and billing finally agree.
Questions?
The 95/5 method is fair when it is documented clearly: measurement interval, direction, CDR, burst policy and how graphs are shared. That is how we approach connectivity at Vortanix—same numbers in the portal, on the invoice and in the looking glass.
Need help reading your traffic graphs or choosing a CDR for a new deployment? Contact our team—we are happy to walk through a sample month with real figures.
Explore related terms in our glossary, or read more articles on the blog home page.