Unmetered bandwidth explained (and why we include it)

Every FlexDedi server includes unmetered traffic: no monthly cap, and no egress or overage fees, inbound or outbound. "Unmetered" means different things in different places, so this post explains the common billing models, what an unmetered port really gives you and how it changes the way you plan costs.

Four ways to bill bandwidth

Quota with overage. The plan includes a monthly allowance, and traffic beyond it is billed per terabyte or slowed down. It is simple, but your bill still depends on a counter you have to watch.

95th percentile. Common in colocation and IP transit. The provider samples your port usage every five minutes, discards the highest 5% of the month's samples and bills the highest remaining one, at a price per Mbps. In a 30-day month that is 8,640 samples, so the 432 discarded samples allow 36 hours of free bursts. The model rewards short spikes and charges for sustained traffic.

Per-GB egress. Typical of public clouds. Every gigabyte that leaves the platform is billed, while inbound traffic is usually free. Cost grows in a straight line with your audience, and you only see the total after the traffic has happened.

Unmetered. You pay for a port of a given speed, and the traffic through it is not counted for billing. This is the FlexDedi model, on ports of 10, 25, 2 × 25 or 2 × 100 Gbps depending on the server.

What an unmetered port means in practice

On an unmetered port, the limit is physics. A port running flat out at 1 Gbps moves 125 MB per second, which is 10.8 TB per day or 324 TB over 30 days. Scaled to our port speeds, the theoretical maximum looks like this:

These figures use decimal units and assume 100% utilization around the clock. Real transfers are lower because of protocol overhead, and servers with 2 × 25 or 2 × 100 Gbps have two such ports.

A useful rule of thumb follows: divide your monthly transfer in terabytes by 324 to get your average rate in Gbps. A service that sends 100 TB a month averages about 0.31 Gbps. Most workloads use a small fraction of a 10 Gbps port on average, so the number to size for is your peak, not your monthly total.

Fair use, honestly

Unmetered pricing only means something if the conditions are clear, so here they are.

• The port is the limit. Using your full port speed around the clock is allowed, and sustained high traffic is not a breach of our acceptable use policy.
• Throughput depends on both ends. Actual throughput to a given destination also depends on networks outside ours, the remote server, TCP tuning and distance. A single TCP stream across an ocean rarely fills a 25 Gbps port, while thousands of viewers together can.
• Normal use covers a lot. Websites, SaaS, streaming, downloads, game servers, CDN nodes and VPN services for your own users are all normal use.
• Two things are excluded. Abuse, such as attacks or spam, is never allowed. Reselling port capacity as IP transit, meaning routing traffic for other networks or selling raw connectivity, requires a prior written agreement.

We include unmetered traffic because we price the port, not the bytes. Once you have chosen a port speed, how much of it you use is up to you.

How it changes cost planning: three examples

The examples below compare traffic costs under per-GB egress with an unmetered port.

Example 1: video streaming

Assume 2,000 concurrent viewers at 5 Mbps each at peak, which is 10 Gbps, and an average of 4 Gbps over the month. At 324 TB per Gbps, that average adds up to 1,296 TB, or 1,296,000 GB. At the assumed $0.05 per GB, egress alone would cost $64,800 a month.

On unmetered ports you size for the peak instead. A 10 Gbps peak needs headroom, so a 25 Gbps server such as the Pro 48 at $999/mo fits, ideally with a second server for redundancy. The traffic itself adds nothing. Our streaming solutions page covers typical setups.

Example 2: nightly off-site backups

Assume you push 2 TB off-site every night, which is 60 TB or 60,000 GB a month. At the assumed rate, that is $3,000 a month.

On a 10 Gbps port, 2 TB takes about 27 minutes at full line rate, or about 33 minutes at a more realistic 8 Gbps, provided the disks at both ends keep up. Restores matter too: under per-GB billing, pulling your data back out is billed again, exactly when you can least afford to wait. A Storage 264 with 264 TB of raw capacity behind a 25 Gbps port makes a natural target; see our storage servers.

Example 3: a public API

Assume 500 requests per second around the clock, with responses of 20 KB on average. That is 10 MB/s, or 80 Mbps, which adds up to 25,920 GB a month: $1,296 at the assumed rate.

On a 10 Gbps port, 80 Mbps is under 1% of capacity, so the server is sized for CPU and memory, and the traffic line disappears from the budget. The bigger benefit is predictability: a tenfold spike from a launch, a scraper or a retry storm changes your load, not your bill.

A planning checklist

1. Measure your current outbound traffic, from your cloud bill or with vnstat -m on existing servers.
2. Convert the monthly total to an average rate by dividing the terabytes by 324, then find your peak in your monitoring.
3. Choose a port with clear headroom above that peak: 10, 25, 2 × 25 or 2 × 100 Gbps.
4. Size CPU, memory and storage for the workload itself, since traffic no longer drives the cost.
5. Plan redundancy across servers or locations for availability, not for bandwidth.

If you are moving away from per-GB billing, our engineers can help you plan the migration at no charge, including how many servers and which port speeds you need.