While everything from cereal boxes to tuna fish cans seem to be shrinking and staying the same price, it looks like Internet access might actually be moving the other direction. On Thursday, the FCC released the second report in its Measuring Broadband America program, which found that ISPs are getting closer to providing consumers the advertised bandwidth they pay for, although America has a long way to go to meet national broadband goals.
The report is an assessment of the bandwidth provided to consumers by major ISPs, broken down by a number of factors including technology, time of day, and performance at different tier levels. A year ago, the FCC found that major U.S. ISPs were getting closer to actually providing the customers the advertised bandwidth for which they were paying. This year, the FCC finds that major ISPs are getting even closer to providing customers with the services they’re advertising — and some providers and technologies regularly exceed advertised performance in some tiers. However, the FCC finds there’s still considerable variation in broadband service, and the U.S. has a long way to go if it’s going to meet the National Broadband Plan goals of getting 50Mbps service to 100 million homes by 2015.
Where are ISPs measuring up and falling down, and how can you tell if your ISP is actually providing you the bandwidth it advertises?
Improvements in 2012
The FCC’s Measuring Broadband America program does not look at every provider and broadband reseller out there on the market, but it does incorporate real performance tests for thousands of subscribers to ISPs that represent about 80 percent of the residential Internet service market for the United States. Although some of the same providers offer commercial Internet services as well, the FCC report focuses purely on the residential broadband experience.
Back in 2011, the FCC generally found the ISPs were delivering about 80 percent of their advertised bandwidth to consumers. Although there was quite a bit of variation across the market (with some providers over-delivering and others under-delivering), on average someone purchasing a package advertising 5Mbps broadband was receiving about 4Mbps of actual service. Reasons for the discrepancies vary widely by technology, provider, and a number of external factors. For DSL, distance from a switch (or “central office”) is a big concern; for cable customers, the number of other subscribers on the same loop can have a major effect — particularly if they’re all putting strains on the system together. (Just think “Netflix prime time.”)
In 2012, the FCC found ISPs are generally more able to deliver their advertised bandwidth to consumers, and the amount of variation between the top and bottom performers shrunk. Where in 2011 there was a 14.4 percent variation between the best and worst performers (across all technologies), in 2012 that dropped to 12.2 percent. In other words, ISPs are about 15 percent better at delivering what they promise. ISPs are also getting better at accurately advertising what they can deliver, instead of promising everyone racing cars and delivering wheel barrows to many folks. They’ve also been significantly investing in their infrastructure to improve performance. As a result, consumers are upgrading, moving from lower broadband tiers to mid- and upper-range broadband where available. Not surprisingly, the upper tiers tend to do a better job of delivering what they promise.
Winners and losers
How did major providers stack up? Here’s the FCC’s year-to-year comparison of observed download speeds to advertised download speeds for fourteen major ISPs:
The numbers show a great deal of progress. The real standout is Cablevision: In 2011 the FCC found that it delivered about 50 percent of advertised bandwidth during peak usage periods. This year, the FCC found it is significantly over-delivering on advertised bandwidth during peak usage periods, offering on average 120 percent of advertised bandwidth. Those are numbers data-hungry Internet fiends like to hear. Verizon’s FiOS fiber service is similarly over-delivering: In 2011 it was offering 114 percent of advertised bandwidth, and this year that’s cranked up to 120 percent. Comcast is also continues to slightly over-deliver on bandwidth to its customers: Last year it was pushing 101 percent of what it was advertising; this year (on average), it’s 103 percent. Mediacom also had a big jump, moving from 75 percent in 2011 to 100 percent in 2012.
Several other companies delivered significant year-to-year improvements on their bandwidth promises, even if they aren’t up to 100 percent. AT&T jumped from 81 to 87 percent, Time Warner moved from 91 to 96 percent, Qwest went from 77 to 83 percent (with new parent company CenturyLink climbing from 87 to 89 percent). Insight, Charter, and Verizon’s DSL services also delivered small improvements.
However, a handful of ISPs slipped from year to year. Frontier actually dropped from delivering 81 percent of advertised bandwidth in 2011 to 79 percent in 2012; Windstream also dropped a point from 85 to 84 percent. And Cox held even, albeit at a pretty respectable 95 percent.
Variations by technology
The FCC notes that broadband providers are generally making major efforts to upgrade their technology and offer both higher tiers and more-reliable service to consumers. But there are some interesting variations in performance based on the type of residential broadband connection in use.
In terms of cumulative sustained download speed, fiber-to-the-home (as typically represented by services like AT&T U-verse and Verizon FiOS) is the clear winner, rarely lower than advertised speeds and frequently offering more bandwidth than providers advertise. Cable comes in second, able to match or exceed advertised bandwidth most of the time. The real loser is DSL, with only about 10 percent of connections able to meet of exceed advertised bandwidth on sustained downloads.
Uploads are a little bit different:
DSL still trails fiber and cable broadband connections for sustained uploads… most of the time. However, about 40 percent of DSL connections meet or exceed advertised upload bandwidth, and at the very high end more DSL connection over-deliver on upload bandwidth than fiber or cable — although the numbers are small.
Another important measure is latency: how long it takes data (on average) to reach its destination so it can be processed. After all, your connection may be able to deliver 40Mbps, but if each packet takes five seconds to get through, it’s still going to feel slow. Again, the FCC finds fiber is the winner on latency, with averages around 20ms across all bandwidth tiers. Cable comes in second, with average latency between 20ms and about 35ms (with better performance generally on higher tiers). While some DSL latency is competitive with cable, the technology generally fares the worst, especially on low-bandwidth connections (512Kbps to about 6Mbps) where the FCC found on average it runs between about 35ms and 55ms.
Bandwidth and congestion
It’s simple to conclude that simply having more bandwidth makes the Web perform faster, and to a certain extent that’s true. Although there are significant variations amongst providers (Cox and Time Warner win basic Web performance in the 1Mbps to 3Mbps broadband tier), subscribers with broadband packages of 10Mbps or greater don’t really see proportionate increases in basic Web performance. This threshold really has nothing to do with ISPs and more to do with the performance of Web browser, the design of typical Web pages, the nature of the HTTP protocol, and the transaction-intensive nature of modern HTML5 and Ajax technologies. Past a certain point, more bandwidth doesn’t translate to a better Web experience, but it does translate to faster downloads and more-reliable bandwidth-intensive services like streaming movies and video chat.
The FCC’s study also found that essentially all broadband technologies experience dips in performance during peak usage hours: Typically, the worse lands between about 8 p.m. and 10 p.m. local time. Although DSL technology is, on average, not likely to deliver full advertised bandwidth, the dip it experiences during peak hours isn’t really much worse than that experienced by cable broadband. Cable (on average) tends to over-deliver during non-peak hours and drop below 100 percent advertised bandwidth during peak usage. Fiber also experiences a peak usage dip in bandwidth, but still consistently over-delivers on advertised capacity.
How to test your own bandwidth
So what does this mean for you?
First and foremost, it means that ISPs are improving on delivering the bandwidth they advertise. High-end fiber-to-the-home services continue to be ahead of the curve, cable is doing well, and DSL still has some catching up to do, but the situation is improving. Of course, every broadband technology has its advantages and drawbacks: Fiber providers typically have no competition in an area and steep bundling requirements, cable means sharing a circuit not just with your local neighbors, and DSL service varies widely with the quality of copper telephone wiring and distance from switches — sometimes it’s great, sometimes not so much.
Want to see how your broadband stacks up? Third-party bandwidth tests are available from sites like DSLReports and Speedtest.net — they can offer some data on how your connection is performing and how it compares to other folks from the same provider and in your area. Be aware the results at any given moment may not be indicative of the average behavior of your connection over time — the tests just take a quick snapshot. Also, some tests perform better than others. For instance, trying a Flash-based bandwidth test on an older machine (say, a PowerPC-based Mac) is going to be a recipe for disappointment: Flash just can’t keep up with high-bandwidth connections. Also remember that simply testing from one machine doesn’t necessarily represent all the devices on your network. If you test your bandwidth on a notebook upstairs while someone is streaming high-definition video using an Xbox downstairs, you may think you’re being ripped off.
The FCC’s Measuring Broadband America reports are one of the only efforts to quantitatively measure how ISPs are measuring up to the promises they make consumers. While the report is only in its second year, it seems to be having a positive impact. After all, the variation amongst ISPs is shrinking, most ISPs improved year to year, and Cablevision in particular seems to have been shamed not just into compliance, but into an outstanding service — at least in terms of delivering advertised bandwidth.
However, the future of Measuring Broadband America report looks a little hazy. Right now, the FCC is collecting some of its data using things like Sam Knows and M-Lab to gather real-world data on residential broadband performance. Telecommunications companies seem to be increasingly keen to shift real-world testing to servers located on their internal networks — a move that has drawn the ire of none other than the “Father of the Internet” Vint Cerf. This could have the effect of both reducing the transparency of the tests (who’s to say the companies won’t fudge things in their favor?) but can also make the tests less indicative of real-world performance, since they’re less likely to have to exchange data with other networks and services. After all, folks who have a “speed test” built into their broadband router know that it almost always passes with flying colors, even when the Internet seems to be moving at a crawl — often because the router’s test is only checking the first link or two in the connectivity chain.
Perhaps more importantly, the ability of ISPs to deliver on bandwidth promises to consumers is only one of many hurdles consumer broadband in America has to overcome. Many Americans are still stuck in areas where there is limited broadband competition (or none at all) thanks to local monopolies and sweetheart deals with local regulators. The FCC’s Measuring Broadband America plan also pretty much only covers wireline access. The goal of using wireless broadband to bring high-speed Internet to rural and under-served areas seems to be on the back burner with the demise of efforts like LightSquared, which means millions of Americans still have little or no choice for broadband Internet…and no hope for that changing anytime soon.