June 1st, 2016
Cisco’s latest VNI Forecast predicts that video will comprise 82% of Internet traffic by 2020; and most of that is entertainment video. It is worthwhile considering what this means for the Internet, and (at least) as a thought experiment, what access networks might look like if the entertainment video and other Internet traffic were carried over separately managed IP networks. This is the question that I am addressing in a paper with Douglas Sicker (CMU) entitled Would You Like Your Internet With or Without Entertainment Video? for presentation at the TPRC in October. The paper hopes to address the deeper underlying question of what should be the optimal convergence path for our digital communications infrastructure, at least with respect to our last-mile access services.
Convergence means different things in technical, economic, and policy terms; and the presumption that “everything over IP,” or even stronger, “everything over the Internet” is efficient, inevitable, or desirable may be wrong.
Building a single network that is optimized for entertainment video traffic might disadvantage other services. Moreover, the economics of media entertainment are distinct from, and potentially in conflict with, the economics of the use cases most often cited as justifying the Internet’s role as essential infrastructure. Finally, separately managing the traffic for Internet and video services may be advantageous in addressing regulatory agenda items such as performance measurement, universal service, online video distribution (OVD) reclassification, and Internet interconnection. While most of the traffic may share the same physical (principally, wired) conduit into homes, it may be more efficient and flexible to segregate traffic into multiple logically distinct networks; and doing so may facilitate technical, market, and regulatory management of the shared resources.
Since most of the traffic on the Internet is already entertainment video, some might regard its convergence with broadband as a fait accompli; however, it remains interesting to consider what the provisioning and investment requirements might be for the other 20%. Some might argue that our inherent uncertainty about what the next “killer app” will be, or even how to characterize future entertainment video, as well as past failures at making predictions, should caution us against trying to forecast future traffic requirements.  Nevertheless, forecast we must and it seems reasonable to accept that entertainment could be a substantially larger load driving peak capacity provisioning decisions depending on how fast entertainment video shifts over-the-top. Moreover, it is worthwhile considering how to provision for the other 20% based on what information we think is available to forecast such traffic, and what to do if some wholly new stream of activity renders our video traffic forecasts gross under-estimates.  Of course, we may simply conclude that the QoS requirements of today’s entertainment video are already sufficiently diverse that building an Internet for such traffic is the best way to build an adequately flexible Internet. Considering these forecasting implications and the underlying assumptions with some care is worthy of further examination.
Others might argue that the marriage between audiovisual entertainment and the Internet is an economic imperative. Access to entertainment media and the associated revenues is what drives consumer broadband adoption and provides the funding for broadband infrastructure investment. Once in place, this broadband infrastructure can support all of the oft-cited economically productive and socially important Internet applications we hope to see in the future (e.g., on-line education, eHealth, smart grids, Internet of Things). Indeed, most of the investment that sustains the Internet today was undertaken to satisfy the demand for ubiquitously available telephone and cable television services – not Internet access. This presumes that alternative funding or business models (e.g., public utility) are neither feasible nor desirable. Even if one accepts the basic premise that entertainment revenues will remain a significant (even dominant) component of the revenue support for shared broadband infrastructure investment, it remains unclear whether content diversity, last-mile competition, or other potential policy goals will be best advanced with full convergence (“everything over the Internet”). For example, the rise of the Internet as an eCommerce platform that allows local retailers to compete in geographically distant markets, also allows distant competitors to compete in local markets, with ambiguous implications for the profitability of local retailers.
Finally, with respect to communications policy, convergence requires us to address the challenge of transitioning from a world of separate regulatory rule-sets for broadcast (TV/mass media), telephone (PSTN), and Internet (which is in the midst of transitioning from mostly deregulated to partially regulated). Each of these rule-sets has sector-specific concerns with little obvious overlap. For example, e911 and lawful surveillance (e.g., wiretaps) have little relevance to legacy broadcast regulation; while program access and copyright are concerns for broadcast regulation with little relevance to legacy PSTN regulation. Furthermore, it is unclear what services and what level of public support is appropriate for infrastructure that is principally used to support high-definition television and other entertainment services. Just as the FCC has chosen to define Broadband Internet Access Service (BIAS) as a new Title II service, it may be feasible for the FCC to define a new Video IP Access Service (VIAS), as a path for segregating the regulatory concerns focused on managing (entertainment) video traffic separately from other BIAS traffic. There are lots of impediments to such a strategy, but further consideration of this or other regulatory options should help illuminate more careful consideration of the implications of convergence, which would be a worthwhile goal even if only as a thought experiment.
In the WEBEX discussions and in the paper, we plan to explore these and related questions and welcome thoughts and comments.
To comment, please “reply all.”
 For example, although today asymmetric, streaming video is the dominant traffic source, it was only a decade ago that peer-to-peer applications like Napster and BitTorrent with more symmetric traffic were dominating aggregate traffic profiles. (However, it is worth noting also that while the mode of delivery was different, entertainment media economics are driving both sources of demand.)
 One the one hand we would have a problem of peak traffic loads growing at much more rapid exponential rates; and on the other, with the “other 20%” shrinking to negligible loads. In the latter case, we have to ask how might we ensure the network remains capable of supporting such traffic (if it is deemed sufficiently socially desirable). This would have technical, economic, and regulatory components.
 Obviously, the entertainment sector is an important nexus for economic activity in its own right; however, it is distinct from many of the other economic sectors (healthcare, education, manufacturing, etc.) that depend on the Internet as a General Purpose Technology (GPT) platform.