Burst rate is the highest speed a service lets you reach above your committed baseline for a limited time. On a network circuit, you might commit to 300 Mbps on a 1 Gbps port and be allowed to burst toward the port speed when traffic spikes. In cloud services, some storage volumes and servers can burst above their baseline IOPS or CPU while they have credits. In each case, the burst is allowed within limits the provider sets, and it does not get the conforming treatment the committed rate does.
At a glance
- Burst rate is a ceiling above your committed rate, not extra committed capacity.
- Providers limit how long or how much you can burst, using token buckets or credit pools.
- Burst traffic is usually the first to be dropped or slowed when the provider’s network is busy.
- Billing varies: bursting can be included, billed as overage, or measured by 95th percentile.
- Burst suits short, occasional peaks; sustained peaks call for a higher committed rate.
What problem it solves
Most business traffic is spiky. Backups, software updates, large file transfers and busy sales days push usage well above normal for minutes or hours, then it falls back. Buying a flat rate sized for the peak means paying for capacity that sits idle most of the month. Burstable services let you commit to, and pay for, a rate that covers normal use, while still letting short peaks through.
The podcast Get More for Less: The Hidden Benefits of Choosing Burst Over Flat Rate Bandwidth covers where this saves money and what to watch in the fine print.
How it works
On network circuits. Dedicated internet access, Ethernet and cloud connection services can be sold with a committed information rate (CIR) below the port speed. Providers commonly enforce the profile with a token bucket. Allowance builds up at the committed rate, and traffic can spend it up to a burst size. When the bucket is empty, traffic above the committed rate is dropped, marked as lower priority, or held back and smoothed by a shaper. Carrier Ethernet services often describe this with a committed rate and burst size plus an excess rate and excess burst size; traffic in the excess band is forwarded only if capacity allows.
In cloud storage and compute. Some disk types and server sizes earn credits while running below baseline and spend them to run faster during spikes. When credits run out, performance drops back to baseline. Workloads that never idle may never build up credits.
On the bill. Some services include bursting in a flat monthly price. Others bill for usage above the commitment, often using 95th percentile billing, which measures usage through the month and ignores the top 5% of samples. Short spikes may cost nothing extra; long ones raise the bill.
Bursting interacts with quality of service (QoS). If all traffic bursts equally, a backup can crowd out a phone call. Prioritizing real-time and business-critical traffic at your edge, and keeping it within the committed rate, gives it the best treatment the service offers when burst headroom runs out. SD-WAN can also move bulk traffic to a second link.
When it matters for buyers
- Sizing a new circuit. Look at a month or more of real traffic: how high the peaks go, how long they last and how often they happen. Commit to the normal busy level; use burst for short, occasional peaks.
- Comparing quotes. A “1 Gbps” quote might be a 1 Gbps commitment or a smaller commitment on a 1 Gbps port. Ask which, and compare total cost under your real traffic pattern, not just the monthly rate.
- At renewal. If you regularly run above your commitment, a higher committed rate may cost less than recurring overages. If you never come close, you may be able to commit to less.
- Protecting voice and video. Real-time traffic should not depend on burst. Keep it inside the commitment and prioritize it.
- Cloud workloads. Check whether disks or servers rely on burst credits, and monitor credit balances so performance doesn’t drop at the worst moment.
Our dedicated internet access page explains how committed and burstable circuits are priced.
Questions to ask vendors
- What is the committed rate, the burst rate (ceiling), and the burst size or credit rules?
- What happens to traffic above the committed rate when your network is busy: dropped, re-marked or delayed?
- Do latency, packet-loss and availability SLAs apply to burst traffic, or only to traffic within the commitment?
- How is burst usage billed: included, overage per Mbps, or 95th percentile? What is the rate?
- Can we raise the committed rate mid-term without restarting the contract?
- Which usage reports will we get to see how often and how long we burst?
How it differs from committed information rate
The committed information rate (CIR) is the rate the provider’s bandwidth profile treats as conforming, and it is what pricing is usually built around. The burst rate is the higher ceiling you may reach above that commitment when capacity allows, for a limited time. Traffic within the CIR generally gets better treatment than burst traffic when the provider’s network is busy, though how much better, and any latency or loss commitments, depend on the provider and must be set out in the SLA or order. A burstable circuit is therefore two numbers, the commitment you can plan on and the ceiling you might reach, and both belong in the contract.
Problems this helps with
- Paying for a big circuit we only fill a few hours a month
- Surprise overage charges after a busy month
- Calls breaking up when backups or updates run
- Not knowing how far above our commitment the circuit will really go
