FCC Broadband Speed Test Finds Improvements Across the Board

But a study released Thursday by the New America Foundation -- which participated in stakeholder talks on the broadband report -- cautioned that U.S. consumers are paying higher prices for slower Internet service when compared to similar cities in other parts of the world. More than 60 academics and engineers told FCC Chairman Julius Genachowski Wednesday they “strongly oppose” a proposal circulating among the FCC and some ISPs to shift away from M-Lab infrastructure. That would “transform an open measurement process into a closed one,” the letter said.

This is the second FCC broadband speed report. The first, released in August, found most ISPs participating in the study were providing, on average, 87 percent of advertised speeds during peak usage periods (CD Aug 3 p2). This week’s report, which measured data collected a little over a year later, found “striking across-the-board improvements on key metrics underlying user performance,” the report said. Speed promises are more accurate, ISPs are more consistently able to deliver advertised speeds, and consumers are subscribing to faster speed tiers.

"Five ISPs now routinely deliver nearly one hundred percent or greater of the speed advertised to the consumer even during time periods when bandwidth demand is at its peak,” the report said. That average download speed during peak usage periods are near 100 percent means consumers today “are experiencing performance more closely aligned with what is advertised than they experienced one year ago,” it said. These improvements were driven by improvements in network performance, not by downward adjustments to the speed tiers offered, the report said.

ISPs more consistently deliver advertised speeds. In 2011, wide variances existed between top and bottom performers in terms of meeting advertised speeds. This year saw a 15 percent reduction in the standard deviation for download speed across DSL, cable and fiber -- meaning ISPs are “doing a better job of delivering what they promise their customers today than they did a year ago,” the report said. The FCC took some credit, saying “there is evidence that our August 2011 Report helped prompt these changes, and had a substantial impact on both the industry and consumer broadband experience."

There is still a fairly large difference in delivered speeds over different technologies. During “peak consumer usage hours” -- defined as weekdays from 7 p.m. to 11 p.m. -- DSL-based services delivered download speeds at 84 percent of advertised speeds, cable-based services delivered 99 percent and fiber-to-the-home services delivered 117 percent. Fiber to the home networks “are out-performing other access technologies, giving consumers consistently higher speed broadband service with lower latency,” the Fiber-to-the-Home Council said. “So much for the ‘conventional wisdom’ that cable broadband performance suffers during peak periods because the network is shared among many customers,” NCTA said (http://xrl.us/bnhgt3).

By ISP, average peak download speeds varied from a high of 120 percent of that advertised to a low of 77 percent. This is a “dramatic” improvement from last year, when Cablevision delivered 54 percent of advertised speed, the report said. The report “demonstrates our commitment to delivering more than 100 percent of the speeds we advertise to our broadband customers,” the company said. The company said it spent $140 million on recent upgrades to its broadband network. Genachowski commended Cablevision at the FCC meeting as “one of this year’s best performers.”

Verizon said the findings “reaffirm the results from the 2011 report, which found that FiOS provides blazing-fast and sustained upstream and downstream speeds even during peak usage periods.” AT&T said the report “demonstrates that consumers continue to get the broadband Internet access speeds they are paying for, that the speeds offered to consumers are increasing and that consumers are moving to these faster broadband speed tiers. Overall, it is abundantly clear that American consumers are getting high-quality broadband services from their Internet Service Providers."

Questions Remain

The New America Foundation isn’t so optimistic on U.S. broadband services compared to the rest of the world. Its Open Technology Institute released a study Thursday comparing high-speed Internet offerings in 22 cities around the world by price, download and upload speed, bundled services, and other metrics, and found that American consumers lag (http://xrl.us/bnhgng). “The Internet download speed Washington DC residents can get for roughly $35 would be over 20 times faster in Hong Kong for around the same price,” NAF said. “And when ranked for their ‘Triple Play’ packages -- Internet, phone, and TV bundles -- Washington, DC comes out way behind other European and Asian cities. For example, residents of Paris pay the equivalent of $35 a month for basic cable TV, phone service, and Internet with download speeds up to 100 mbps.” The U.S. needs to re-examine its current policies, and rather than focus on spectrum auctions and the promise of wireless broadband, “policymakers need to address the lack of competition in most of the U.S. and how policies can enable new competitors to enter the marketplace,” the report said.

Genachowski said the higher price point for lower broadband speeds compared to densely populated countries around the world is “an issue that we need to continue to make progress on in the US.” But “the gap is closing,” and the FCC is continuing to promote competition “using all the different tools we have,” Genachowski said in response to our question at a news conference. He pointed out areas where the U.S. leads, such as the widescale deployment of 4G LTE services, and mobile innovation, leading the world in the apps economy, he said.

As the commission continues monitoring Internet speeds and other metrics, commissioners and officials from the Office of Engineering and Technology repeatedly promise a commitment to openness and transparency. But dozens of prominent engineers and academics wrote to Genachowski Wednesday to express concern over a proposal they said would “replace the Measurement Lab server infrastructure with closed infrastructure,” run by the participating ISPs whose own speeds are being measured in the commission’s broadband transmission tests (http://bit.ly/Q8cn9J). “We strongly oppose any decision by the FCC to run a closed measurement program,” said the letter whose signers included Internet pioneer Vint Cerf of Google and NAF Vice President Sascha Meinrath. “For the scientific process to work, measurement data must be openly available as well as access to methodologies, and explicit cataloging of assumptions is essential if results are to be confirmed and replicated. A switch from an open to a closed infrastructure makes this process impossible or, at best, questionable."

FCC officials rejected the assertion they want to “replace” M-Lab’s servers. “That statement is false,” said Chief Technology Officer Henning Schulzrinne. “We're not considering replacement of the M-Lab infrastructure. We have enjoyed working with them.” The proposal would add redundancy by installing more servers so it becomes easier to detect anomalies, Schulzrinne said: There have been “discussions to enhance -- but not replace."

The proposal would change the way “whiteboxes” test collected data, said a copy of the plan we obtained that’s from participants in the FCC’s test. Whiteboxes now test their data against an “off-net” pool of servers -- servers that measure packets traveling across the public Internet, that are provided solely by M-Lab -- and an “on-net” pool, provided solely by the ISP, which measures data that has not yet left the network and gone onto the public Internet. The proposal would allow the off-net pool of servers be either the “best research platform server,” run by M-Lab or some other entity, or “best ‘public’ ISP-provided server."

This would shift the measurement away from M-Lab’s open platform to measurement servers controlled by the ISPs themselves, said NAF’s Ben Lennett, a representative of the lab. Redundancy and the collection of more data “makes sense at a high level, but we have offered to integrate the ISPs’ donated servers” into the M-Lab infrastructure, and ISPs haven’t taken the offer, he said. “The question is really about the credibility of the measurement.” M-Lab is completely open with how it runs its servers, but “there’s no indication that that is going to be the case with ISP-controlled servers,” Lennett said. Walter Johnston, chief of the FCC Electromagnetic Compatibility Division, told us any data collected by ISPs as part of the official commission broadband speed test would be made public, and it would be clear who gathered the data.