Last week Cisco Systems released its latest Visual Networking Index report (0). The white paper is part of an ongoing series that tracks and forecasts trends in regional and global Internet Protocol (IP) traffic. As is often the case, the report contains a staggering amount of information about the contours of the global internet. But the findings are especially interesting because of their implications for tech policy.
Smartphones are the future
The report relays some important findings about the devices and networks we use to access the internet. In 2016, wired devices accounted for a bare majority (51 percent) of IP traffic. But by 2021, that figure will fall to 37 percent, compared to 63 percent for mobile and Wi-Fi devices. More astoundingly, smartphone traffic will soon pass personal computer (PC) traffic: By 2021, smartphones will account for a third of all IP traffic, while PCs will comprise only a quarter.
This shift toward wireless networks, which is the culmination of long-term trends, suggests that the premium that some elites have placed on fiber buildout is misplaced. Fiber optic cable is still the gold standard for data transfer and will remain important for wireless backhaul. But if Cisco’s projections are correct, policymakers and pundits should focus at least as much on ways to encourage wireless buildout. Reforms such as removing local barriers to construction, making rural subsidies more flexible, and promoting new avenues such as millimeter wave spectrum (1) will be important to meet the coming shift and should occupy a greater mindshare among telecom elites than gigabit wired networks.
Rise of the content delivery networks
For me, perhaps the most surprising finding is the supremacy of content delivery networks (CDNs) in the internet ecosystem. CDNs are privately-built alternatives to the public internet. By 2016, more than half of all IP traffic crossed a CDN on its way to its destination. By 2021, that figure will rise to a staggering 71 percent.
This trend has significant implications for the future of telecom policy. The net neutrality debate has sucked all the oxygen from the room for more than a decade, but of course, net neutrality rules do not apply to CDNs. This raises the specter that, like the 1996 Telecommunications Act regarding the wired telephone network and payphones, we may exhaust ourselves getting network regulation right, only to watch traffic sidestep those networks in favor of more efficient models.
Of course, the analogy is not perfect, as CDNs often rely on traditional broadband providers for last-mile delivery. But this suggests interconnection disputes between the networks that comprise the internet will play an increasing role going forward. Unfortunately, regulators know little about the interconnection market. What we do know is that CDNs compete against traditional peering arrangements and transit networks in an environment marked by falling prices and cutthroat margins, suggesting that competition is fierce and that run-of-the-mill business disputes should not immediately trigger the impulse to regulate. Pundits should focus more on understanding the contours of this space as the spotlight shifts from the edge and moves toward the core of the network.
The Cisco report contains one interesting finding relevant to that inquiry: 68 percent of CDN traffic (or just under half of all IP traffic worldwide, if my math is correct) will be carried by private CDNs by 2021. In contrast to “traditional” CDNs, private CDNs are built and operated by content providers, such as Netflix, Google, Amazon, and Facebook, for their own content and do not make excess capacity available for purchase by rival content providers. This suggests that the internet ecosystem is experiencing greater vertical integration. This is not, by itself, a bad thing, as vertical agreements often benefit consumers. But those pundits worried about greater vertical integration by broadband providers should be willing to examine similar behavior by edge providers at the other end of the internet ecosystem.
Video remains the killer app
For all our pontificating about the internet as an unparalleled tool for public debate, education, and personal enrichment, it appears that most of us are experiencing it primarily as the “new television.” IP video already dominates all rivals, comprising 73 percent of all IP traffic globally. Cisco estimates this volume will grow threefold by 2021, to constitute 82 percent of all IP traffic.
That’s a lot of cat videos.
Consumers’ increasing reliance on IP video as a substitute for traditional broadcast and cable content has significant implications for the shape of IP traffic. For example, Cisco shows that peak-time traffic grew at a much faster rate than overall IP traffic, and it projects that this growth rate disparity will continue, driven by the growth of video as a “prime time” phenomenon and by demand for real-time video, which has a higher peak-to-average ratio than on-demand video. This suggests that network congestion, which has largely disappeared from the public debate with the 4G revolution, may reemerge as a policy issue.
Overall, these trends suggest that Federal Communications Commission (FCC) Chairman Ajit Pai is correct to seek more data-driven policy recommendations. The FCC is placing greater emphasis on economics, to correct for the perception that the Open Internet Order’s conclusions reflected a dearth of economic input. The Cisco report reinforces a point often made by former TechPolicyDaily blogger Richard Bennett (2): The agency should seek similar input from the engineers who actually run the networks and understand these trends in real time.