Last week FCC Chairman Tom Wheeler announced a new strategy to drive Net neutrality, laying out a plan to require open, unconstrained Internet access, get rid of last-mile tariffs and unbundling, and eliminate any throttling or charging for class of service. This time, as he wrote in a Wired opinion piece, the proposal is to use the FCC's Title II authority to recast Internet service providers as common carriers and avoid having the courts throw out the ruling.
While many agree with the concept of open access, others argue that not being able to charge for access will constrain deployment of new infrastructure. I will leave the arguments about the value of Net neutrality to others, focusing here instead on how an open, unconstrained Internet could impact any of us delivering real-time communications systems.
The Downside of Coexistence
Without throttling or prioritization, our real-time communications will have to coexist with a wide range of traffic that has other demands, both in terms of latency as well as loss. That begs the question: Are we raising the potential of an Internet that will not accommodate real-time traffic? (For more Net neutrality insight, read Any New Fire in the Latest Neutrality Smoke?)
As I detailed last week on No Jitter, VoIP is already beginning to suffer from a variety of issues (see VoIPmageddon: Is Quality Leading to a Telephony Meltdown?). Taking away the ability to prioritize real-time communications in the open Internet (or even within networks) could exacerbate those.
As an early VoIP adopter, I suffered through my fair share of quality problems. In the early days, I remember experiencing dramatic loss of quality on my VoIP calls from home, via a cable modem, at 3 p.m. when kids came home from school. Presumably their use of early file-sharing applications like Napster caused the performance hit. Today, Comcast prioritizes voice traffic, but will it be able to do that under Wheeler's Net neutrality ruling? Likewise, will voice over LTE providers be able to prioritize voice traffic over video downloads from Netflix?
Doomed if We Don't
Quite frankly, I believe we do need a way to prioritize real-time traffic -- or suffer the consequences. If we are not able to segregate real-time communications we could see a dramatic reduction in quality. And that will add to the other issues that are already degrading voice quality.
The saving grace is that real-time communications is a small portion of traffic on both corporate LANs and the Internet. On a typical gigabit LAN, real-time traffic often accounts for less than a few percent of the total. On these networks, simple class-of-service features like absolute priority and never discard assure that the real-time traffic moves freely while any reduction of total bandwidth is irrelevant for other services and applications. (For more details on this, read my white paper, "CoS and QoS - Managing Bandwidth, Complexity, and Cost.)
The key question is: Can we apply the same logic to the Internet?
A Fixed Allotment
I believe that guaranteeing availability of 10% of any Internet capacity for priority traffic would be sufficient -- today, tomorrow, and forever. In fact, the vast majority of the traffic increase on the Internet is entertainment, which is not real-time and can be buffered. The 10% rule would be simple and could be applied everywhere.
In this scenario, if you purchased 10 megabits per second (Mbps) of inbound bandwidth from your Internet access provider, then you would be entitled to mark up to 1 Mbps of traffic as a "real-time class" (RTC). Your ISP would then carry this traffic as absolute priority, never discard, assuring it moves at essentially the root performance of the network and is not impacted by any other traffic. All other traffic would never have access to less than 90% of the total available bandwidth, assuring that that traffic can move freely.
Similarly, when ISPs peer, they would be required to accept up to 10% of the peering bandwidth as the RTC class, providing the same service level for that traffic as it traverses their networks. The result is that the users make the decision on classifying up to 10% of their purchased bandwidth to be in the RTC class, assuring that that traffic will move freely and at speeds and losses essentially at the "best" performance of the set of links through which the traffic passes.
The beauty of the 10% RTC scheme is that it puts the requirement of classification on every Internet entry point. So your router, whether home or enterprise, would soon be capable of marking your real-time traffic as RTC, and could prioritize different data types should your RTC traffic volume exceed your 10% capacity. (Alternatively, depending on the frequency of this happening, you might buy more bandwidth overall). However, once an RTC packet enters the Internet, it gets priority handling along the full path.
Companies like Google or Netflix could have 10% of this inbound traffic be RTC, but not all. They would have the same rules applied to their traffic entering the network as you; RTC is limited to 10% of purchased inbound traffic capacity. I assume this would lead to great innovation in how to use specialized marking mechanisms to optimize the speed of playback, starting with sending some traffic using RTC and migrating to "normal" (non-RTC) marked traffic as the flow progresses and gets cached. This would allow rapid changes to be sent with defined timing, while bulk viewing traffic would not get RTC treatment.
Hopefully, as the FCC considers the new proposal for Net neutrality and as other political bodies get involved, everyone will consider the needs of the real-time market and not just streaming and other one-way services. VoIP, video, WebRTC, and a wide variety of other real-time services depend on the Internet to deliver the quality that users and businesses demand. We cannot let a race to equality ignore the practical needs of real-time services.
Please, Mr. Chairman and the rest of the FCC, think about giving us 10%! We really need an Internet that is not only open, but also can support real-time communications.
Join Phil Edholm at Enterprise Connect Orlando in EC Summit: Managing in a Software-Intensive World. Save $300 on conference passes when you register today with discount code: NJSPEAKER.
