Measuring Broadband America is a U.S. government-sponsored program managed by the Federal Communications Commission (FCC). It was developed to assess U.S. consumers’ Internet services. The MBA program enlists volunteers across the entire country to volunteer to attach a simple test agent, also referred to as the ‘Whitebox’, to the home Internet router.
Since 2010, the FCC has been managing the MBA program to better understand the nation’s broadband Internet performance.
This program is the only one of its kind in terms of its open and transparent approach to measurement, data analysis, and data sharing. In particular, its collaborative approach between Internet Service Providers (ISPs), academics, government agencies, and all interested parties is unique. Feedback from all, especially the public, is welcomed as it enables the program to continually improve and evolve.
Each household’s data is anonymized and aggregated to present a country-wide perspective.
The success of Measuring Broadband America (MBA) is thanks to the contributions of many consumers, academics, industry veterans, standards bodies, Internet companies and research institutions who have all given up their time and expertise. Each participant is required to sign up to the program’s principles which define a Code of Conduct. The entire measurement methodology of the MBA is published, as is the data, to maintain the highest possible levels of transparency.
For ISPs, the Code of Conduct sets out guiding principles on how to interact with the data to maintain the integrity of the study and its results. Volunteers do not sign the Code of Conduct but do sign an End-user License Agreement (EULA). This document outlines what their data will be used for as well as other terms and conditions that are in place when taking part in the study. This must be agreed to before a volunteer is sent a Whitebox.
Academics wishing to use the MBA platform for experiments and research under the MBA-Assisted Research Study (MARS) program again have a separate Academic Code of Conduct. While we sincerely welcome the contribution of all stakeholders, we are always mindful of the extremely high standards of the initiative.
The following ISPs actively participate in the MBA program. If you subscribe to an Internet service with one of these ISPs, you may be able to contribute to the program.
We are also looking for subscribers from AT&T, Hughes, and Viasat as part of the recruitment initiative. If you subscribe to Internet services from these providers you may be able to contribute to the program.
The ISPs included in the 2019 sample plan are:
ATT, Centurylink, Charter, Cincinnati Bell, Comcast, Cox, Frontier, Hawaiian Telecom, Mediacom, Optimum, Verizon, Windstream
The 2020 sample plan will be created in March 2020 in preparation for the September 2020 reporting month.
The MBA project is open to anyone with an Internet connection. Since the first MBA report in August 2011, over 300,000 American citizens have volunteered from all across the United States. If you would like to sign up, all you need to do is register here and confirm which ISP you’re with and what speed tier you purchase from them.
When selected, you will receive a Whitebox to connect to your router. The Whitebox runs a series of performance tests and reports these test results back to the FCC.
If you are a volunteer for the MBA program, you will receive some valuable benefits. You will be able to see your own test results using SamKnows One, our cloud-based platform, and track your ISP's performance over time. By having a reporting Whitebox, you will receive detailed performance data so you can understand your personal Internet connection and monitor your network health.
Once the sample plan has been created, the next step is to ensure a good number of homes are online in order to include their results in the final report.
Volunteers hear about the project via:
It is possible for anyone to sign up at any point during the year.
It’s important that the speed tier information for volunteers is correct. There are multiple ways of verifying a volunteer’s speed tier. Initial throughput tests are used to confirm reported speeds once customers sign up (to confirm they have provided the correct speed tier information). The broadband service tier reported by each panelist is also validated in the following way:
The following four steps are completed for each panelist:
Where volunteers are excluded (e.g if they have a business Internet connection or are on a speed tier that is not on the sample plan) this is noted and published on the FCC’s website.
The sample plan outlines which ISPs and speed tiers will be measured in each reporting year. This document is updated every year to make sure that the project measures the most popular speed tiers available from each ISP. The participating ISPs are the largest in the USA. New speed tiers are regularly introduced or upgrades made so it’s important that the document is dynamic. We consider the most popular speed tiers to be the ones that have 80% of an ISP's subscriber base on them - meaning the results are relevant to the largest base of people possible.
We are working towards being able to measure every ISP in the USA and always looking for new sign ups. If you would like to get involved, sign up here.
The MBA Core Panel consists of Whiteboxes, small hardware devices that plug straight into your home router and test ISP performance. The Whiteboxes are sent to volunteers that sign up and are on an ISP and speed tier that is in the sample plan. Once plugged in the Whiteboxes run a suite of active measurements related to network performance. This data is used in the MBA yearly reports.
SamKnows maintains a large (650,000+) control panel of routers which are also in homes across America. These routers execute a very similar set of measurements to the FCC's Measuring Broadband America program.
The measurements from the SamKnows Control Panel run to a separate set of test servers, which are not used by the FCC's Measuring Broadband America program. The measurements are executed according to a test schedule, in a similar fashion to the Whiteboxes, and provide us with an extremely large set of download speed, upload speed, latency, and packet loss results.
There are currently over 650,000 homes in the SamKnows Control Panel.
The Whiteboxes in consumers' homes carry out their speed measurements against test servers in nine cities across the US. These test servers are deployed at neutral peering locations, split across three separate hosting providers, and are connected to major Internet exchanges, typically at 10Gbps or greater.
Additionally, some ISPs opt to also install test servers inside their network, to provide an additional comparison of 'on-net' versus 'off-net' performance. The results from ISP-hosted test servers are not used in the Measuring Broadband America reports.
Only results from the off-net test servers, deployed at the neutral peering locations across the US, are used in the creation of the Measuring Broadband America reports.
The MBA program has policies that must be fulfilled in order for a hosting provider’s test servers to be used in the project. These are split in 5 categories that outline important considerations for the project.
Yes. ISPs are able to challenge data or analysis in the published reports. An ISP raising a concern follows the process outlined below. The ISP raises an issue with the FCC's report, outlining their concern or challenge. Once this has been done, the following process is followed:
The creation of each Measuring Broadband America report involves a significant review process to ensure the validity and robustness of the data being reported. This can lead to some data being excluded.
Whiteboxes can be excluded from a Measuring Broadband America report for the following reasons:
Additionally, measurement data can be excluded on specific dates for the following reasons:
Please note that regardless of whether data is excluded from the report, the raw data exports (published on the FCC website) always include all the data.
We recognize that running a fully transparent program carries the risk that individuals may attempt to 'game' the system.
This is why we maintain so many checks of the data. For example, data from the Core Panel is verified against data from the Control Panel. The ISPs have no knowledge of where these routers are. The Control Panel uses entirely different test servers. There is no way whatsoever for an individual to game the data of the Control Panel.
The data from the Control Panel is used to check against the data from the Core Panel.
All raw data collected as part of the MBA project is published and available for academics to review and report on. Academics are also encouraged to use the MBA test platform for experiments in order to further the understanding of how the internet operates. Academics wishing to run a project can submit their proposal to the MBA-Assisted Research Study (MARS) team for consideration. If approved, these experiments can run across the MBA platform and the data and analysis is subsequently published to help further the advancement of specific technical areas of network research.
The following academic institutions have collaborated with the MARS team to run experiments on the MBA platform:
The process for research organizations works in a similar way to academic institutions. The MBA-Assisted Research Studies (MARS) scheme allows research institutions to utilize either the vast amount of data collected or the existing test platform to run network experiments. The proposals are discussed collectively at collaborative meetings and then rolled out when the MBA platform is not in use for a reporting period. These results are collected and published in scientific publications. These are also shared with the collaborative during the monthly meetings that occur.
The following research institutions have collaborated with the MARS team to run experiments on the MBA platform:
It is very easy for academic institutions to get involved with the MBA project. Measuring Broadband America collects data 24 hours a day, 7 days a week on a variety of metrics. As part of the FCC’s commitment to openness and transparency, academic institutions have access to all the data collected in the program. This can be found on the FCC’s website here.
Academics are also able to suggest experiments to run on the MBA platform, which fall under the MBA-Assisted Research Studies (MARS) cover projects. These are distinct from the routine reporting and MBA measurement operations aimed at producing year-on-year reports of network performance. As their name suggests, MARS projects mostly consist of standalone experiments initiated due to requests from research institutions engaged in advancing the “state of the art” in network measurements. In a few cases, this component of the MBA program also incorporates pilot efforts which require particularly close technical coordination across FCC teams.
Before a MARS project can begin, the following steps are required:
1. The teams involved develop a project proposal describing the experiment or feature amendment proposed, and the associated intended use of the MBA infrastructure, for review within the FCC.
2. In the cases of projects in categories A and B:
All information is available on the FCC’s MARS page found here.
A variety of studies have been undertaken on the MBA test platform over the years, these are outlined on the FCC’s MARs page here and below:
1. Locating Last-Mile Performance Bottlenecks
Dr. Sri Sundaresan and Professor Nick Feamster, Georgia Institute of Technology/Princeton University
2. Calibrating geolocation algorithms for use in the Time Series Latency Probe (TSLP) and other projects, Drs. kc Claffy/Brad Huffaker, Cooperative for Applied Internet Data Analysis(CAIDA), UCSD.
3. Highly Granular Packet Loss Data collection.
This is packet loss data aggregated by the minute instead of the hour as normally done in the Fixed Broadband Measurements. This was undertaken at the request of a team consisting of researchers from MIT CSAIL (Drs. David Clark, Steve Bauer) and CAIDA/UCSD (Drs. kc Claffy, Amogh Dhamdhere)
4. NAT Revelio: Detecting NAT 444 in the ISP
This is a project led by Drs. Amogh Dhamdhere and kc. Claffy at the Cooperative for Applied Internet Data Analysis(CAIDA), University of California, San Diego (UCSD), partnering with Simula Research, Norway and the University of Carlos III, Madrid, Spain. As indicated in the title, it investigates one of the means of stretching now exhausted IPv4 address space in the evolving Internet. These experiments were done in two phases. The publication links are below.
5. Updated data consumption metric
A ‘data consumption’ or ‘data usage’ metric has been in operation as part of the fixed MBA test suite since the start of the MBA program in 2011.
The measurement device in use in the fixed MBA program (referred to elsewhere in MBA documentation as the ‘Whitebox’) is based on a hardware platform positioned within the home network so as to be able to perform the required measurement functions. Over the years since the program’s inception, both the ongoing evolution in consumer Internet services, and the associated modifications to connection methods and on-premise equipment offered by ISPs to their customer base, have necessitated accompanying changes to the relative positions and functions of the ‘Whitebox’ within MBA panel participants’ home networks.
As a result, the initial version of the data usage metric remained viable on a progressively smaller fraction of the MBA test panel, in which the ‘Whiteboxes’ (in addition to their measurement function) were also configured to provide the operative in-home Wi-Fi Access points. The proportion of Whitebox installations where this configuration was instantiated had dwindled considerably since 2011.
In response to recurring requests from multiple quarters, in mid-2018, the data consumption metric was updated so it could operate uniformly across the entire MBA measurement panel regardless of the specifics of Whitebox configuration. The metric was further enhanced so it can distinguish the relative proportions of Internet-related Wi-Fi and wired traffic. The existing capability to measure the volume of test-generated traffic was retained and all constituent measurements within the metric (see data dictionary link below ) were designed to correlate in terms of measurement interval and protocol layer.
The data dictionary for the updated usage metric is available here: 2018 usage data dictionary.
A detailed technical description can be found here: Technical Methodology for measuring data usage.
The datasets collected so far (to be accrued per month) from deploying this metric are at the following:
Yes. Anyone running tests on the MBA platform must sign the good faith agreement that outlines how they will interact with the platform to protect the integrity of the data collected. This is shared here.
All of the data that is collected is published and made publicly available on the FCC’s website here. Generic top level data on performance metrics is also available under license. The image below shows a sample of an ISP's performance over the 9 years of Measuring Broadband America. This shows average performance increasing from around 27Mbps in 2011 to just over 140Mbps.
Alternative filters include: ISP, Speed Tier, Unit ID (a unique identifier for the Whitebox only), Agent, Device Type, Country, IP Version, Package download speed, Package Upload speed, Peak vs. Off Peak,Peak and 24/7, Region, State, Target, Target Group, Technology.
Use cases include the FCC tracking how performance is impacted by extreme weather events such as hurricanes. All data is completely anonymized and no personally identifiable information is shared. This is to protect the identity of the volunteers and the location of the devices.
The MBA panel consists of many thousands of SamKnows enabled devices. Whiteboxes are installed in volunteer’s homes and run a suite of tests 24 hours a day, 7 days a week. In addition we have over 100,000 SamKnows enabled routers that use the SamKnows SDK as well as Netgear routers with the SamKnows SDK enabled also. These SamKnows enabled routers act as our control panel to verify our speed test data. We use the Whiteboxes as the dedicated volunteer panel for collecting these measurements and running all the tests.
The Code of Conduct is a good faith agreement any entity who takes part in the project must sign. The code sets out guiding principles for how stakeholders will interact with the data and helps ensure the integrity of the study and its results. The Code of Conduct is published in the technical appendix here.
There is a subtlety in the maximum speeds of the Whiteboxes, and these have different behaviours on measurements. Here are two example scenarios:
The first scenario is where the maximum measurable speed of a Whitebox is limited by the network interface. For example, the 100Mbps limit of the WR741ND(v4) is caused by the 100Mbps network interfaces on the Whitebox. It actually has a powerful enough processor to transfer data to around 200Mbps, but it never gets the chance because the ports are the constraining factor. In this scenario, if you deploy this Whitebox on a good-performing 200Mbps connection then you expect to get measurements flat-lining at ~95Mbps (100Mbps minus packet overheads).
The second scenario is where the maximum measurable speed is limited by the processing power of the Whitebox, and not the network interfaces. For example, the WNR3500L has 1Gbps network ports, but doesn't have a fast enough processor to transfer data at 1Gbps. In this scenario, you do not get a nice stable result at 150Mbps (our advised limit for the WNR3500L). You will instead get erratic and unpredictable behaviour. Sometimes it will measure far beyond 150Mbps, and sometimes not.
The erratic behaviour beyond the advised limit is caused by the CPU (Central Processing Unit) hitting 100% utilisation. When this occurs, the device can't process any more packets, which ultimately leads to packet loss. And because we're testing over TCP, packet loss causes the flow of data to slow down (this is by design in TCP - the receiver signals to the sender "You're sending data too fast, I can't keep up, please slow down!"), which leads to lower and more volatile results.
So whilst it looks counterintuitive at first, it is entirely possible that a legacy Whitebox with a 150Mbps limit which is deployed on a 200Mbps tier could measure less than 150Mbps.