Measuring Fixed Broadband - Eleventh Report

Eleventh
Measuring Broadband America
Fixed Broadband Report

A Report on Consumer Fixed Broadband Performance
in the United States

Federal Communications Commission

Office of Engineering and Technology

Table of Contents

• Executive Summary
• Summary of Key Findings  
  • Most popular advertised service tiers
  • Median download speeds
  • Variations in speeds
  • Latency
  • Packet loss
  • Web Browsing Performance
• Methodology  
  • Participants
  • Measurement process
  • Measurement tests and performance metrics
  • MBA data coverage and quality
• Test Results  
  • Most popular advertised service tiers
  • Observed Median download and upload Speeds
  • Variations in speeds
  • Latency
• Additional test results  
  • Actual speed, by service tier
  • Variations in speed
  • Web browsing performance, by service tier
• Annex: MBA Performance During COVID-19 Pandemic

1. Executive Summary

The Eleventh Measuring Broadband America Fixed Broadband Report (“Eleventh Report” or “Report”) contains validated data collected in September and October 2020^[1] from fixed Internet Service Providers (ISPs) as part of the Federal Communication Commission (FCC)’s Measuring Broadband America (MBA) program.  This program is an ongoing rigorous nationwide study of consumer broadband performance in the United States.  The goal of this program is to measure the network performance realized on a representative sample of service offerings and the residential broadband consumer demographic across the country.^[2]   This representative sample is referred to as the MBA “panel.” Thousands of volunteer panelists are drawn from the subscriber bases of ISPs, which collectively serve a large percentage of the residential marketplace.

The initial Measuring Broadband America Fixed Broadband Report was published in August 2011^[3] and presented the first broad-scale study of directly measured consumer broadband performance throughout the United States.  Including this current Report, eleven reports have now been issued.^[4]   These reports provide a snapshot of fixed broadband Internet access service performance in the United States utilizing a comprehensive set of performance metrics.  As part of an open data program, all methodologies used in the program are fully documented, and all collected data is published for public use without any restrictions.  The collected performance data can be used for various additional studies and analysis.  

For the period covered by this Eleventh Report, the coronavirus pandemic had a significant impact on each of the metrics tracked by the Measuring Broadband America program.  A national emergency was declared in the United Sates on March 13, and by the time the validated data collection period for this Report was underway, approximately 42% of the U.S. workforce was working remotely from their homes.^[5]   Similarly, 93% of households reported some form of remote learning and distance education during this time period.^[6]   The 2020 data would reflect these developments, as well as the impacts of various measures taken by the Federal Communications Commission and other federal actors to manage the crisis and support recovery.  Despite all the challenges, the MBA measurements showed that for the most part, the ISP networks were able to keep up with this unprecedented increased load on their networks, as will be evidenced by the results presented in this report

A. Major Findings Of The Eleventh Report

The key findings of this report are:

• The maximum advertised download speeds amongst the service tiers offered by ISPs and measured by the FCC ranged from 100 Mbps to 1,000 Mbps for the period covered by this report.
• The weighted average advertised speed of the participating ISPs was193.9 Mbps, representing an increase of33% from the previous year’s Tenth Report and over 43% from the Ninth Report.
• For most of the major broadband providers that were tested, measured download speeds were 100% or better than advertised speeds during the peak hours (7 p.m. to 11 p.m. local time).
• Ten ISPs were evaluated in this report.  Of these, CenturyLink, Cincinnati Bell, and Frontier employed multiple different broadband technologies across the U.S, whereas the rest were single technology providers.  Overall, 12 different ISP/technology configurations were evaluated in this report; nine performed at or better than their advertised speed during the peak hours.  One performed below 90% for actual-to-advertised download speed during the peak hours.
• In addition to providing download and upload speed measurements of ISPs, this report also provides a measure of consistency of measured-to-advertised speeds of ISPs with the use of our “80/80” metric.  The 80/80 metric measures the percentage of the advertised speed that at least 80% of subscribers experience at least 80% of the time over peak periods. By this measure, seven of the 12 ISP/technology configurations performed at or better than 95% of their advertised speed during the peak hours. Two of the 12 ISP/technology configurations provided lower than 75% of advertised speed to at least 80% of panelists for at least 80% of the time.

These and other findings are described in greater detail within this report.

B. Speed Performance Metrics

Speed (both download and upload) performance continues to be one of the key metrics reported by the MBA.  The data presented includes each ISP’s broadband performance as the median^[7] of speeds experienced by panelists within a specific service tier.  These reports mainly focus on common service tiers used by an ISP’s subscribers.^[8]

Additionally, consistent with previous Reports, we also compute average per-ISP performance by weighting the median speed for each service tier by the number of subscribers in that tier.  Similarly, in calculating the composite average speed, taking into account all ISPs in a specific year, the median speed of each ISP is used and weighted by the number of subscribers of that ISP as a fraction of the total number of subscribers across all ISPs.

In calculating these weighted medians, we draw on two sources for determining the number of subscribers per service tier.  ISPs may voluntarily contribute subscription demographics per surveyed service tier as the most recent and authoritative data.  Many ISPs have chosen to do so.^[9]   When such information is not provided by an ISP, we instead rely on the FCC’s Form 477 data.^[10]   All facilities-based broadband providers are required to file data with the FCC twice a year via Form 477 regarding deployment of broadband services, including subscriber counts.  For this report, we used the June 2019 Form 477 data.  

As in our previous reports, we found that, for most ISPs (except for a few DSL providers and one fiber provider), the measured speeds experienced by subscribers either nearly met or exceeded advertised service tier speeds.  However, since we started the MBA program, consumers have changed their Internet usage habits.  In 2011, consumers mainly browsed the web and downloaded files; thus, we reported mean broadband speeds since these statistics were likely to closely mirror user experience.   Consumer internet usage has now become dominated by video consumption, with consumers regularly streaming video for entertainment and education^[11]   Our network performance analytics were expanded to use consistency in service metrics to better capture the shift in usage patterns.  Using both median measured speed and consistency in service metrics better reflects the consumer’s perception and usefulness of Internet access service given this shift in usage habits.

We use two kinds of metrics to reflect the consistency of service delivered to the consumer:  First, we report the percentage of advertised speed experienced by at least 80% of panelists during at least 80% of the daily peak usage period (“80/80 consistent speed” measure).  Second, we show the fraction of consumers who obtain median speeds greater than 95%, between 80% and 95%, and less than 80% of advertised speeds.

C. Use Of Other Performance Metrics

Although download and upload speeds remain the network performance metric of greatest interest to the consumer, we also spotlight two other key network performance metrics in this report: latency and packet loss.  These metrics can affect significantly the overall quality of Internet applications.

Latency is the time it takes for a data packet to travel across a network from one point on the network to another.  High latencies may affect the perceived quality of some interactive services such as phone calls over the Internet, video chat and video conferencing, or online multiplayer games.  All network access technologies have a minimum latency that largely is determined by the data rate limitations of the technology. There are many other factors that affect latency though, including the location of the server with which one is communicating, the route taken to the server, and whether or not there is any congestion on that route.

Packet loss measures the fraction of data packets sent that fail to be delivered to the intended destination.  Packet loss may affect the perceived quality of applications that do not incorporate retransmission of lost packets, such as phone calls over the Internet, video chat, certain online multiplayer games, and certain video streaming services.  During network congestion, both latency and packet loss typically increase.  High packet loss degrades the achievable throughput of download and streaming applications.  However, packet loss of a few tenths of a percent, for example, is common and is unlikely to affect significantly the perceived quality of most Internet applications.

The Internet continually evolves in its architecture, performance, and services.  Accordingly, we will continue to adapt our measurement and analysis methodologies to further improve the collective understanding of performance characteristics of broadband Internet access.  By doing so we aim to help the community of interest across the board, including consumers, technologists, service providers, and regulators.

2. Summary of Key Findings

A.Most Popular Advertised Service Tiers

A list of the ISP download and upload speed service tiers that were measured in this report are shown in Table 1.  It should be noted that while upload and download speeds are measured independently and shown separately, they typically are offered by an ISP in a paired configuration.  The service tiers that are included for reporting represent the top 80% (therefore ‘most popular’) of an ISP’s set of tiers based on subscriber numbers.  Taken in aggregate, these plans serve the majority of the subscription base of the ISPs across the USA.

Table 1: List of ISP service tiers whose broadband performance was measured in this report

*Tiers that lack sufficient panelists to meet the program’s target sample size.

**  Tiers near Gigabit speeds whose technical methodologies for reporting have not yet been established for the MBA program.  We note that a Technical Working Group including providers of these tiers has been created to evaluate different test methodologies for reporting on results for 940 Mbps and above download speeds.

Chart 1.1 (below) displays the weighted (by subscriber numbers) mean of the top 80% advertised download speed tiers for each participating ISP for the last four years (September 2017 to September-October 2020) grouped by the access technology used to offer the broadband Internet access service (DSL, cable, or fiber).  It should be noted that this chart only aims to represent the weighted average of the ISP’s advertised speeds as one index to track broadband service evolution; it does not reflect the measured performance of the ISPs. As far as a further composite index across the entirety of the ISPs in the MBA sample,  in September-October 2020, the weighted average advertised download speed was 193.9 Mbps among the measured ISPs, which represents a 33% increase compared to the average, in September-October 2019, of 146.1 Mbps (as in the Tenth MBA Report) and a 166% increase compared to the  2017 of 72.9 Mbps^[12] (as shown in the Eighth MBA Report).

Chart 1.1: Weighted average advertised download speed among the top 80% service tiers offered by each ISP

Nearly all of the ISPs had higher weighted averages of advertised speeds in September-October 2020 as compared to September-October 2019.  

Chart 1.1 reflects that  DSL speeds lag far behind the speed of other technologies.  In order to better compare the DSL speed offerings by the various ISPs, we have added a separate Chart 1.2 drawn to a scale that makes their relative speeds more apparent.

Chart 1.2: Weighted average advertised download speed among the DSL ISPs

As can be seen in Chart 1.1, there is considerable difference between the offered average weighted speed tier by technology.  Chart 2 plots the weighted average of the top 80% ISP tiers by technology for the last four years.^[13]  All technologies showed increases in the set of advertised download speeds by ISPs.  For the September-October 2020 period, the weighted mean advertised speeds for DSL technology was 21 Mbps, which lagged considerably behind the weighted mean advertised download speeds for cable and fiber technologies of 178 Mbps and 447 Mbps, respectively.  DSL technology speed increased from 13 Mbps to 21 Mbps from 2019 to 2020, a 60% increase.  In comparison, cable technology showed an 15% increase in weighted mean advertised download speed from 2019 to 2020. The largest increase in weighted average advertised speeds, however, was for Fiber technology, which showed a 114% increase from 2019 to 2020.

Chart 2: Weighted average advertised download speed among the top 80% service tiers based on technology.

Chart 3 plots the migration of panelists to a higher service tier based on their access technology.^[14]   Specifically, the horizontal axis of Chart 3 partitions the September-October 2019 panelists by the advertised download speed of the service tier to which they were subscribed.  For each such set of panelists who also participated in the September-October 2020 collection of data,^[15] the vertical axis of Chart 3 displays the percentage of panelists that migrated by September-October 2020 to a service tier with a higher advertised download speed.  There are two ways that such a migration could occur: (1) panelist may have changed their broadband plan during the intervening year to a service tier with a higher advertised download speed, or (2) a panelist the panelist’s ISP increased the advertised download speed of the panelist’s existing subscribed plan.^[16]

Chart 3 shows that the percentage of panelists subscribed in September-October 2019 who moved to higher tiers in September-October2020 was between 3% to 25% for DSL subscribers, 5% to 58% for cable subscribers, and 5% to 74% for fiber subscribers.  Interestingly, a large number of subscribers, between 30% to 60%, migrated to a higher speed tier using a different technology from what they had in September-October2019.This is in sharp contrast to the previous year when only 1% to 8% of subscribers migrated to a different technology. It is possible that with the increased reliance on use of home networks for performing school and work-related activities during the COVID-19 pandemic, many consumers may have migrated away from older and slower speed technologies such as DSL. ^[17]

Chart 3: Consumer migration to higher advertised download speeds

B. Median download speeds

Advertised download speeds may differ from the actual speeds that subscribers experience.  Some ISPs meet network service objectives more consistently than others; some meet such objectives unevenly across their geographic coverage area.  Also, speeds experienced by a consumer may vary during the day if the aggregate user demand during busy hours causes network congestion.  Unless stated otherwise, the data used in this report is based on measurements taken during peak usage periods, which we define as 7 p.m. to 11 p.m. local time.^[18]

To compute each ISP’s overall performance inclusive of all sampled tiers, we determine the ratio of the median speed for each tier to the advertised tier speed and then calculate the weighted average of these based on the subscriber count per tier.  Subscriber counts for the weightings were provided from the ISPs themselves or, if unavailable, from FCC Form 477 data.

Chart 4 shows the ratio of the measured median download and upload speeds experienced by an ISP’s subscribers to that ISP’s advertised download and upload speeds weighted by the subscribership numbers for the tiers.  The actual speeds experienced by the subscribers of most ISPs are close to or exceed the advertised speeds.  DSL broadband ISPs, however, continue to advertise “up-to” speeds that, on average, exceed the actual speeds experienced by their subscribers.  Out of the 12 ISP/technology configurations shown, 9 met or exceeded their advertised download speeds, and two more reached at least 90% of their advertised download speed.  Only Cincinnati Bell DSL (at 78%) performed below 90% of its advertised download speed.

The upload speed performance followed a similar trend to the download speed performance, with 7 of the 12 ISP/technologies exceeding their advertised upload speed, and  2 falling below 90% of the advertised upload speed.

Chart 4: The ratio of weighted median speed (download and upload) to advertised speed for each ISP.

C. Variations In Performance/Measured speeds

As discussed earlier, actual speeds experienced by individual consumers may vary by location and time of day.  Chart 5 shows, for each ISP, the percentage of panelists who experienced a median download speed (averaged over the peak usage period during our measurement period) that was greater than 95%, between 80% and 95%, or less than 80% of the advertised download speed.

Chart 5: The percentage of consumers whose median download speed was greater than 95%, between 80% and 95%, or less than 80% of the advertised download speed

Between 3% to 75% of subscribers of ISPs using DSL technology experienced speeds greater than or equal to 95% of their respective ISP’s advertised download speeds during peak hours.  ISPs using cable technology and fiber technology provided speeds equal to or better than 95% of their advertised speeds to between 89%-98% and 53%-98% of their subscribers, respectively.

Though the median download speeds experienced by most ISPs’ subscribers nearly met or exceeded the advertised download speeds, there were a number of customers of each ISP for whom the median download speed fell significantly short of the advertised download speed. The best performing ISPs, when measured by this metric, were Charter, Comcast, Cox, Mediacom, Optimum, Frontier-Fiber and Verizon-Fiber; more than 88% of their panelists experienced an actual median download speed of at least 95% of the advertised download speed.

In addition to variations based on a subscriber’s location, speeds experienced by a consumer may fluctuate during the day.  This typically is caused by increased traffic demand and the resulting stress on different parts of the network infrastructure.  To examine this aspect of performance, we use the term “80/80 consistent speed.”  This metric is designed to assess temporal and spatial variations in measured values of a user’s download speed.^[19]  While consistency of speed is in itself an intrinsically valuable service characteristic, its impact on consumers will hinge on variations in usage patterns and needs.  As an example, a good consistency of speed measure is likely to indicate a higher quality of service experience for internet users consuming video content.

Chart 6 summarizes, for each ISP, the ratio of 80/80 consistent median download speed to advertised download speed, and, for comparison, the ratio of median download speed to advertised download speed shown previously in Chart 4.  The ratio of 80/80 consistent median download speed to advertised download speed was less than the ratio of median download speed to advertised download speed for all participating ISPs due to congestion periods when median download speeds were lower than the overall average.  The size of the difference between these two ratios is indicative of the variability of median download speed: when the difference between these two ratios is small, the median download speed is fairly insensitive to both geography and time; in contrast, when the difference between the two ratios is large, there is a greater variability in median download speed, either across a set of different locations or across different times during the peak usage period at the same location.

Chart 6: The ratio of 80/80 consistent median download speed to advertised download speed.

As can be seen in Chart 6, cable and fiber ISPs performed better than DSL ISPs with respect to their provision of consistent speeds.  Customers of Charter, Comcast, Cox, Mediacom, Optimum, Frontier Fiber, and Verizon, for example, experienced median download speeds that were very consistent; i.e., they provided greater than 95% of the advertised speed during peak usage period to more than 80% of panelists for more than 80% of the time. In particular, Charter, Comcast, Cox and Mediacom provided 100% or greater than their advertised speed during the peak usage period to more than 80% of their panelists for more than  80% of the time. In contrast, the 80/80 consistent download speed for Cincinnati Bell DSL was 46% of the advertised speed.

D.Latency

The latency between any two points in the network is the time it takes for a packet to travel from one point to the other.  It has a fixed component that depends on the distance, the transmission speed, and transmission technology between the source and destination, and a variable component due to queuing delay that increases as the network path congests with traffic.  The MBA program measures latency by measuring the round-trip time between the consumer’s home and the closest measurement server.

Chart 7 shows the median latency for each participating ISP.  In general, higher speed service tiers have lower latency, as it takes less time to transmit each packet.  The measured median latencies ranged from 8 ms to 28 ms (with the exception of CenturyLink DSL, for which the median latency measure was 37 ms).

Chart 7: Latency by ISP

Measured DSL latencies (between 21 ms to 37 ms) were slightly higher than those for cable (12 ms to 26 ms), and latencies were lowest for fiber ISPs (8 ms to 13 ms). 

E. Packet Loss

Packet loss is the percentage of packets that are sent by a source but not received at the intended destination.  The most common causes of packet loss are congestion leading to buffer overflows or active queue management along the network path.  Alternatively, high latency might lead to a packet being counted as lost if it does not arrive within a specified window.  A small amount of packet loss is expected, and indeed packet loss commonly is used by some Internet protocols such as TCP to infer Internet congestion and to adjust the sending rate to mitigate the offered load, thus lessening the contribution to congestion and the risk of lost packets.  The MBA program uses an active UDP-based packet loss measurement method and considers a packet lost if it is not returned within 3 seconds.

Chart 8 shows the average peak-period packet loss for each participating ISP, separated into three groups based on technology used for the access network.  The average peak-period packet loss performance is represented in three bands based on commonly accepted packet loss thresholds for specific services and typical packet loss limits in provider Service Level Agreements (SLAs) for enterprise services (as consumer offerings typically are not subject to SLAs).  Separating the packet loss performance into these three bands allows a more granular view of the packet loss performance of the ISP network.  The breakpoints for the three bands are: up to 0.4% packet loss, 0.4% - 1% packet loss, and over 1% packet loss.  The 1% standard for packet loss is commonly accepted as the point at which highly interactive applications such as VoIP experience significant degradation in quality according to industry publications and international (ITU) standards.^[20]  The specific value of 0.4%, roughly the middle ground between the highly desirable performance of 0% packet loss described in many documents for Voice of Internet Protocol (VoIP), generally is supported by major ISP SLAs for network.  Indeed, most SLAs support 0.1% to 0.3% packet loss guarantees,^[21] but these are generally for enterprise level services which entail business-critical applications that require some service guarantees.

Chart 8:  Percentage of consumers whose peak-period packet loss was less than 0.4%, between 0.4% to 1%, and greater than 1%.

Chart 8 shows that ISPs using fiber technology have the lowest packet loss, and that ISPs using DSL technology tend to have the highest packet loss.  As shown in this chart, 4% to 8% of DSL subscribers experience 1% or greater packet loss.  The corresponding numbers for cable and fiber are 0% to 5% and 0% to 0.6%, respectively.  Within a given technology class, packet loss also varies among ISPs. It should be noted that most ISPs use fiber for their backbone irrespective of the access technology they may deploy.

F. Web browsing performance

The MBA program also conducts a specific test to gauge web browsing performance.  The web browsing test accesses nine popular websites that include text and images, but not streaming video.  The time required to download a webpage depends on many factors, including the consumer’s in-home network, the download speed within an ISP’s network, the web server’s speed, congestion in other networks outside the consumer’s ISP’s network (if any), and the time required to look up the network address of the webserver.  Only some of these factors are under the control of the consumer’s ISP.  Chart 9 displays the average webpage download time as a function of the advertised download speed.  As shown by this chart, webpage download time decreased as download speed increased, from about 11 seconds at 1.5 Mbps download speed to about 1.9 seconds for 25 Mbps download speed.  Subscribers to service tiers exceeding 25 Mbps experienced slightly smaller webpage download times decreasing to 1.4 seconds at 175 Mbps.  Beyond 175 Mbps, the webpage download times decreased only by minor amounts. These download times assume that only a single user is using the Internet connection when the webpage is downloaded; our testing does not account for additional common scenarios, such as where multiple users within a household are simultaneously using the Internet connection for viewing web pages, as well as other applications such as real-time gaming or video streaming.

Chart 9: Average webpage download time, by advertised download speed.

3. Methodology

A. Participants

Eleven ISPs actively participated in the Fixed MBA program in September-October 2020.   They were:

• Altice Optimum
• CenturyLink
• Charter Communications
• Cincinnati Bell
• Comcast
• Cox Communications
• Frontier Communications Company
• Hawaiian Telecom ^[22]
• Mediacom Communications Corporation
• Verizon
• Windstream Communications

The methodologies and assumptions underlying the measurements described in this Report are reviewed at meetings that are open to all interested parties; these meetings are documented in public ex parte letters filed in the GN Docket No. 12-264.  Policy decisions regarding the MBA program were discussed at these meetings prior to adoption, and involved issues such as inclusion of tiers, length of test periods, mitigation of operational issues affecting the measurement infrastructure, and terms of use for notification to panelists.  Participation in the MBA program is open and voluntary.  Participants include members of academia, consumer equipment vendors, telecommunications vendors, network service providers, consumer policy groups, and the FCC’s contractor for this project, SamKnows.  In 2020-2021, participants at these meetings (collectively and informally referred to as “the FCC MBA broadband collaborative”), included all eleven participating ISPs and the following additional organizations:

• AT&T
• Level 3 Communications (“Level 3”), now part of CenturyLink
• Massachusetts Institute of Technology (“MIT”)
• StackPath
• NCTA – The Internet & Television Association (“NCTA”)
• New America Foundation
• Princeton University
• United States Telecom Association (“US Telecom”)
• University of California - Santa Cruz
• University of Chicago

Participants have contributed in important ways to the integrity of this program and have provided valuable input to FCC decisions for this program.  Initial proposals for test metrics and testing platforms were discussed and critiqued within the broadband collaborative.   We thank all the participants for their continued contributions to the MBA program.

B. Measurement process

The measurements that provided the underlying data for this report were conducted between MBA measurement clients and MBA measurement servers.  The measurement clients (i.e., whiteboxes) were situated in the homes of 5,951 panelists, each of whom received service from one of the 10 evaluated ISPs.  The evaluated ISPs collectively represent over 70% of U.S. residential broadband Internet connections.  After the measurement data was processed (as described in greater detail in the Technical Appendix), test results from 2,488 panelists were used in this report^[24] .

The measurement servers used by the MBA program were hosted by StackPath, and Level 3 Communications, and were located in thirteen cities (often with multiple locations within each city) across the United States near a point of interconnection between the ISP’s network and the network on which the measurement server resided.

The measurement clients collected data throughout the year, and this data is available as described below.  However, only data collected from September 2 – 24, 2020 (inclusive) plus September 26 – October 2, 2020 (inclusive), referred to throughout this report as the “September-October 2020 reporting period,” were used to generate the charts and conclusions in this Report.^[25]

Broadband performance varies with the time of day.  At peak hours, more people tend to use their broadband Internet connections, giving rise to a greater potential for network congestion and degraded user performance.  Unless otherwise stated, this Report focuses on performance during peak usage period, which is defined as weeknights between 7:00 p.m. to 11:00 p.m. local time at the subscriber’s location.  Focusing on peak usage period provides the most useful information because it demonstrates what performance users can expect when Internet access services in their local area experience the highest demand from users.  As shown in the Annex to this Report, it should be noted that the peak period remained the same despite the increased number of people working from home due to the pandemic.

Our methodology focuses on the performance of each of the participating ISP’s access network.  The metrics discussed in this Report are derived from active measurements, i.e., test-generated traffic flowing between a standalone measurement client, located within a panelist’s home, and a measurement server, located outside the ISP’s network.  For each panelist, the tests make a periodically-rerun empirical choice as to the measurement server that has the lowest latency to the measurement client.  Thus, the metrics measure performance along the path followed by the measurement traffic within each ISP’s network, through a point of interconnection between the ISP’s network and the network on which the chosen measurement server is located.  The service performance that a consumer experiences, however, could differ from our measured values for one or more reasons.

First, as noted above, in the course of each test instance, performance is measured only to a single measurement server rather than to multiple servers. This is consistent with the approach chosen by most network measurement tools, although applications may vary.  As a point of comparison, the average web page may load its content from a multiplicity of end points.

In addition, bottlenecks or congestion points in the full path traversed by consumer application traffic also impact a consumer’s perception of Internet service performance.  These bottlenecks may exist at various points: within the ISP’s network, beyond its network (depending on the network topology encountered en route to the traffic destination), in the consumer’s home on the Wi-Fi used to connect to the in-home access router, or from a shortfall of capacity at the far end point being accessed by the application.  The MBA tests explore how a service performs from the point at which a fixed ISP’s Internet service is delivered to the home on fixed infrastructure (deliberately excluding Wi-Fi, due to the many confounding factors associated with it) to the point at which the test servers are located.  As MBA tests are designed to focus on the access to the ISP’s network, they will not include phenomena at most interconnection points or transit networks that consumer traffic may traverse.

Since the MBA program is focused on ISP network performance, and the overall performance a consumer experiences with their service can be affected by congestion that may arise at other points in the path potentially taken by consumer traffic (e.g., in-home Wi-Fi, peering points, transit networks, etc.), the MBA methodologies/system/measurements are designed to recognize and eliminate such confounding factors outside an access ISP’s control.^[26]   A consumer’s home network, rather than the ISP’s network, may be a bottleneck with respect to congestion.  We measure the performance of the ISP’s service delivered to the consumer’s home network, but this service often is shared simultaneously among multiple users and applications within the home.  In-home networks, which typically include Wi-Fi, may not have sufficient capacities to support peak loads.^[27]

In addition, consumers’ experience of ISP performance is manifested through the set of applications they utilize.  The overall performance of an application depends not only on the network performance (i.e., raw speed, latency, or packet loss), but also on the application’s architecture and implementation and on the operating system and hardware on which it runs.  While network performance is considered in this Report, application performance is not.  To the extent possible, the MBA focuses on performance within an ISP’s network.

C. Measurement Tests And Performance Metrics

This Report is based on the following measurement tests:

• Download speed: This test measures the download speed of each whitebox over a 10-second period, once per hour during peak hours (7 p.m. to 11 p.m.) and once during each of the following periods: midnight to 6 a.m., 6 a.m. to noon, and noon to 6 p.m. The download speed measurement results from each whitebox are then averaged across the measurement month, and the median value for these average speeds across the entire set of whiteboxes on a given tier is used to determine the median measured download speedfor that tier. The overall ISP download speed is computed as the weighted median for each service tier, using the subscriber counts for the tiers as weights.
• Upload speed: This test measures the upload speed of each whitebox over a 10-second period, which is the same measurement interval as the download speed. The upload speed measured in the last five seconds of the 10-second interval is retained, the results of each whitebox are then averaged over the measurement period, and the median value for the average speed taken over the entire set of whiteboxes is used to determine the median upload speed for a service tier. The ISP upload speed is computed in the same manner as the download speed.
• Latency and packet loss: These tests measure the round-trip times for approximately 2,000 packets per hour sent at randomly distributed intervals. Response times less than three seconds are used to determine the mean latency. If the whitebox does not receive a response within three seconds, the packet is counted as lost.
• Web browsing: The web browsing test, performed once every hour, measures the total time it takes to request and receive webpages, including the text and images, from nine popular websites. The measurement includes the time required to translate the web server name (URL) into the webserver’s network (IP) address.

This Report focuses on three key performance metrics of interest to consumers of broadband Internet access service, as these metrics are likely to influence how well a wide range of consumer applications work: download and upload speed, latency, and packet loss.  Download and upload speeds  also are the primary network performance characteristic advertised by ISPs.  However, as discussed above, the performance observed by a user in any given circumstance depends not only on their ISP’s network, but also on the performance of other parts of the Internet and of the application platform  used. The Technical Appendix to this Report describes each test in more detail, including additional tests not utilized for the purpose of compiling  this Report.

D. MBA data coverage and quality

The MBA panel sample used in the reporting period is validated (i.e., upload and download tiers of the whiteboxes are verified with providers), and the measurement results are carefully inspected to eliminate any statistical outliers.  This leads to a ‘validated data set’ that accompanies each report.  The Validated Data Set^[29] on which this Report is based, as well as the full results of all tests, are available at http://www.fcc.gov/measuring-broadband-america.  For interested parties, as tests are run 24x7x365, we also provide raw data (referred to as such because validation cross-checks are not done except in the test period used for the report, thus subscriber tier changes may be missed) for the reference month and the rest of the year.  Previous reports of the MBA program, as well as the data used to produce them,  also are available at the site listed above.

Both the Commission and SamKnows, the Commission’s contractor for this program, recognize that, while the methodology descriptions included in this document provide an overview of the project, interested parties may be willing to contribute to the project by reviewing the methodology used in the testing.  SamKnows welcomes review of its methodology and technical platform, consistent with the Commission’s goals of openness and transparency for this program.^[30]