Category Archives: Internet Traffic

How can U.S. broadband lag if it generates 2-3 times the traffic of other nations?

Is the U.S. broadband market healthy or not? This question is central to the efforts to change the way we regulate the Internet. In a short new paper from the American Enterprise Institute, we look at a simple way to gauge whether the U.S. has in fact fallen behind other nations in coverage, speed, and price . . . and whether consumers enjoy access to content. Here’s a summary:

  • Internet traffic volume is an important indicator of broadband health, as it encapsulates and distills the most important broadband factors, such as access, coverage, speed, price, and content availability.
  • US Internet traffic — a measure of the nation’s “digital output” — is two to three times higher than most advanced nations, and the United States generates more Internet traffic per capita and per Internet user than any major nation except for South Korea.
  • The US model of broadband investment and innovation — which operates in an environment that is largely free from government interference — has been a dramatic success.
  • Overturning this successful policy by imposing heavy regulation on the Internet puts one of America’s most vital industries at risk.

Interconnection: Arguing for Inefficiency

Last week Level 3 posted some new data from interconnection points with three large broadband service providers. The first column of the chart, with data from last spring, shows lots of congestion between Level 3 and the three BSPs. You might recall the battles of last winter and early spring when Netflix streaming slowed down and it accused Comcast and other BSPs of purposely “throttling” its video traffic. (We wrote about the incident here, here, here, and here.)

The second column of the Level 3 chart, with data from September, shows that traffic with two of the three BSPs is much less congested today. Level 3 says, reasonably, the cause for the change is Netflix’s on-net transit (or paid peering) agreements with Comcast and (presumably) Verizon, in which Netflix and the broadband firms established direct connections with one another. As Level 3 writes, “You might say that it’s good news overall.” And it is: these on-net transit agreements, which have been around for at least 15 years, and which are used by Google, Amazon, Microsoft, all the content delivery networks (CDNs), and many others, make the Net work better and more efficiently, cutting costs for content providers and delivering better, faster, more robust services to consumers.

But Level 3 says despite this apparent improvement, the data really shows the broadband providers demanding “tolls” and that this is bad for the Internet overall. It thinks Netflix and the broadband providers should be forced to employ an indirect A–>B–>C architecture even when a direct A–>C architecture is more efficient.

The Level 3 charts make another probably unintended point. Recall that Netflix, starting around two years ago, began building its own CDN called OpenConnect. Its intention was always to connect directly to the broadband providers (A–>C) and to bypass Level 3 and other backbone providers (B). This is exactly what happened. Netflix connected to Comcast, Verizon, and others (although for a small fee, rather than for free, as it had hoped). And it looks like the broadband providers were smart not to build out massive new interconnection capacity with Level 3 to satisfy a peering agreement that was out of balance, and which, as soon as Netflix left, regained balance. It would have been a huge waste (what they used to call stranded investment).

Twitch Proves the Net Is Working

Below find our Reply Comments in the Federal Communications Commission’s Open Internet proceeding:

September 15, 2014

Twitch Proves the Net Is Working

On August 25, 2014, Amazon announced its acquisition of Twitch for around $1 billion. Twitch  ( is a young but very large website that streams video games and the gamers who play them. The rise of Twitch demonstrates the Net is working and, we believe, also deals a severe blow to a central theory of the Order and NPRM.

The NPRM repeats the theory of the 2010 Open Internet Order that “providers of broadband Internet access service had multiple incentives to limit Internet openness.” The theory advances a concern that small start-up content providers might be discouraged or blocked from opportunities to grow. Neither the Order nor the current NPRM considers or even acknowledges evidence or arguments to the contrary — that broadband service providers (BSPs) may have substantial incentives to promote Internet openness. Nevertheless, the Commission now helpfully seeks comment “to update the record to reflect marketplace, technical, and other changes since the 2010 Open Internet Order was adopted that may have either exacerbated or mitigated broadband providers’ incentives and ability to limit Internet openness. We seek general comment on the Commission’s approach to analyzing broadband providers’ incentives and ability to engage in practices that would limit the open Internet.”

The continued growth of the Internet, and the general health of the U.S. Web, content, app, device, and Internet services markets — all occurring in the absence of Net Neutrality regulation — more than mitigate the Commission’s theory of BSP incentives. While there is scant evidence for the theory of bad BSP behavior, there is abundant evidence that openness generally benefits all players throughout the Internet value chain. The Commission cannot ignore this evidence.

The rise of Twitch is a perfect example. In three short years, Twitch went from brand new start-up to the fourth largest single source of traffic on the Internet. Google had previously signed a term sheet with Twitch, but so great was the momentum of this young, tiny company, that it could command a more attractive deal from Amazon. At the time of its acquisition by Amazon, Twitch said it had 55 million unique monthly viewers (consumers) and more than one million broadcasters (producers), generating 15 billion minutes of content viewed a month. According to measurements by the network scientist and Deepfield CEO Craig Labovitz, only Netflix, Google’s YouTube, and Apple’s iTunes generate more traffic.


U.S. Share of Internet Traffic Grows

Over the last half decade, during a protracted economic slump, we’ve documented the persistent successes of Digital America — for example the rise of the App Economy. Measuring the health of our tech sectors is important, in part because policy agendas are often based on assertions of market failure (or regulatory failure) and often include comparisons with other nations. Several years ago we developed a simple new metric that we thought better reflected the health of broadband in international comparisons. Instead of measuring broadband using “penetration rates,” or the number of  connections per capita, we thought a much better indicator was actual Internet usage. So we started looking at Internet traffic per capita and per Internet user (see here, here, here, and, for more context, here).

We’ve update the numbers here, using Cisco’s Visual Networking Index for traffic estimates and Internet user figures from the International Telecommunications Union. And the numbers suggest the U.S. digital economy, and its broadband networks, are healthy and extending their lead internationally. (Patrick Brogan of USTelecom has also done excellent work on this front; see his new update.)

If we look at regional comparisons of traffic per person, we see North America generates and consumes nearly seven times the world average and more around two and a half times that of Western Europe.

Looking at individual nations, and switching to the metric of traffic per user, we find that the U.S. is actually pulling away from the rest of the world. In our previous reports, the U.S. trailed only South Korea, was essentially tied with Canada, and generated around 60-70% more traffic than Western European nations. Now, the U.S. has separated itself from Canada and is generating two to three times the traffic per user of Western Europe and Japan.

Perhaps the most remarkable fact, as Brogan notes, is that the U.S. has nearly caught up with South Korea, which, for the last decade, was a real outlier — far and away the worldwide leader in Internet infrastructure and usage.

Traffic is difficult to measure and its nature and composition can change quickly. There are a number of factors we’ll talk more about later, such as how much of this traffic originates in the U.S. but is destined for foreign lands. Yet these are some of the best numbers we have, and the general magnitudes reinforce the idea that the U.S. digital economy, under a relatively light-touch regulatory model, is performing well.

Arbor’s new Net traffic report: “This is just the beginning…”

See this comprehensive new Web traffic study from Arbor Networks — “the largest study of global Internet traffic since the start of the commercial Internet.” 


Internet is at an inflection point

Transition from focus on connectivity to content
Old global Internet economic models are evolving
New entrants are reshaping definition / value of connectivity

New technologies are reshaping definition of network
“Web” / Desktop Applications, Cloud computing, CDN

Changes mean significant new commercial, security and engineering challenges

This is just the beginning…

These conclusions and the data Arbor tracked and reported largely followed our findings, projections, and predictions from two years ago:

And an update from this spring:

Also see our analysis from last winter highlighting the evolution of content delivery networks — what my colleague George Gilder dubbed “storewidth” back in 1999 — and which Arbor now says is the fastest growing source/transmitter of Net traffic.

Jackson’s traffic spike

Om Malik surveys the Net traffic spike after Michael Jackson’s death:

Around 6:30 p.m. EST, Akamai’s Net Usage Index for News spiked all the way to 4,247,971 global visitors per minute vs. normal traffic of 2,000,000, a 112 percent gain.