Measurement methodology and technical docs
The full reference for how the browser test runs, what is scored, what is exported, and what the result page deliberately leaves out.
knowledge base
Internet reliability, explained
This is the learning hub for Bufferbloat.org: short guides, measurement notes, and research references about what makes an internet connection feel reliable in real life. Bufferbloat is the starting point, but the bigger question is connection quality under actual use. For why the project exists at all, read the mission.
start with your line
Run the test, then use the guides to understand the result.
The scorecard gives the rest of this page a concrete reference: quiet-line ping, ping while download and upload traffic are active, throughput, p95 spread, and an overall reliability grade.
article paths
Choose the question you are trying to answer
These articles are meant to be complementary. Start with the practical question, then follow the links inside each guide when you need the measurement details.
Start here
The core reference pages behind the scorecard.
What a bufferbloat test actually measures
The best first read after running the test: throughput, quiet-line ping, loaded ping, p95 spread, and why this is not just another speed test.
What internet connection quality really means
A broader guide to why one number cannot explain whether a line feels usable, unreliable, stable, or frustrating in daily use.
Why latency under load matters
The central bufferbloat concept: ping can look fine when the line is quiet, then change when download or upload traffic starts.
How to inspect and export your test data
What the technical details drawer contains, what the CSV export is for, and where the exact field definitions live.
Measurement concepts
Why the scorecard uses these specific measurement choices.
Why we use latency spread, not jitter
Why p95 spread is more useful than average jitter or worst ping for spotting repeated high-delay moments during real load.
Why we use median ping, not average ping
Why the middle of the measured samples is a better reference point for a browser-based reliability test than the arithmetic average.
Topics this knowledge base should cover next
I want to cover the topics people actually run into: why throughput is not the whole meaning of speed, why fast internet can feel slow, why slower connections can still feel usable, how Wi-Fi and routers change the result, and what fixes are realistic at home.
Real-life use cases
Use these when the problem is how the connection feels.
Internet reliability test
Start here if the line feels unpredictable and ordinary tests do not explain why calls, browsing, or shared use still suffer.
Internet stability test
Use this when the connection feels jumpy: fine one moment, then suddenly delayed, spiky, or inconsistent under normal use.
Internet latency test
A guide to ping, response time, and why a quiet latency number is not enough to judge real-life network quality.
Calls internet test
Why video and audio calls can fail on connections that have enough throughput, especially while uploads or other devices are active.
Zoom is right: “Your internet connection is unstable”
Bufferbloat.org is not affiliated with Zoom; this explains the network behavior that makes meeting-like traffic degrade under load.
How to read video-meeting test results
After running the test, use upload load, download load, quiet ping, and latency spread to understand meeting reliability.
Gaming network test
Low-latency games expose delay and ping spread quickly, even when the connection looks fast in a conventional test.
bibliography
Technical references and further reading
These are sources I respect and consider foundational to this work: community projects, standards, algorithms, and tools that shaped how Bufferbloat.org thinks about bufferbloat, queue management, loaded latency measurement, and practical fixes in real networks.
Bufferbloat Project wiki
A readable entry point for symptoms, experiments, fixes, glossary material, and the broader history of bufferbloat research.
ACM Queue: BufferBloat
The 2011 ACM Queue discussion with Vint Cerf, Van Jacobson, Nick Weaver, and Jim Gettys that helped bring the problem to wider attention.
RFC 7567: Active Queue Management
IETF recommendations for active queue management, the class of techniques used to control queueing delay before buffers fill.
RFC 8289: CoDel
The RFC for Controlled Delay Active Queue Management, one of the foundational approaches to controlling bufferbloat-generated delay.
RFC 8290: FQ-CoDel
The flow-queueing CoDel scheduler and AQM algorithm, published by authors from the bufferbloat community.
CAKE on Bufferbloat.net
Documentation for CAKE, a queue management and traffic-shaping system designed for home gateways and last-mile connections.
OpenWrt
Open-source router firmware commonly used with SQM, CAKE, and FQ-CoDel to reduce loaded latency on home networks.
Flent
The FLExible Network Tester, widely used for repeatable network experiments such as RRUL and loaded-latency testing.
M-Lab NDT
Measurement Lab's Network Diagnostic Tool and public data platform for internet performance research.
Cloudflare Speed Test
A public browser-based network performance test that reports throughput, latency, latency spread, and related metrics.
Waveform Bufferbloat Test
A popular consumer bufferbloat test and useful point of comparison for explaining loaded latency to a broader audience.
LibreQoS
Open-source quality-of-experience tooling for ISPs, closely aligned with modern queue management and latency-aware networks.