FCC NPRM prohibits good network management
The new FCC NPRM regulations on broadband providers in its current draft form will prohibit good network management because it bans any sort of discrimination in favor of or against any particular applications. That means real-time applications like online gaming or Voice over IP (VoIP) which typically require less than 100 Kbps of bandwidth (4% of the total capacity provided by a 3 Mbps Broadband service) will be forced to suffer from aggressive applications. More specifically, NPRM paragraph 104 would prohibit any kind of discrimination even if it’s reasonable.
NPRM Paragraph 104 – “Based on the record, we propose a general rule prohibiting a broadband Internet access service provider from discriminating against, or in favor of, any content, application, or service, subject to reasonable network management”
Paragraph 137 in the NPRM debates whether exceptions should be made for real-time applications like VoIP and online gaming and it is an open question at this point.
NPRM Paragraph 137 – “Some have suggested it would be beneficial for a broadband provider to protect the quality of service for those applications for which quality of service is important by implementing a network management practice of prioritizing classes of latency-sensitive traffic over classes of latency-insensitive traffic (such as prioritizing all VoIP, gaming, and streaming media traffic). Others have suggested that such a practice would be difficult to implement in a competitively fair manner and could undermine the benefits of a nondiscrimination rule, including keeping barriers to innovation low.”
If paragraph 137 is permitted to stand in its current state, the NPRM will regulate good network management out of existence.
Why prioritization is needed for real-time applications
VoIP or online gaming applications need to transmit and receive tiny 223-byte packets every 20 milliseconds and if the packets suddenly take an additional 50 to 200 milliseconds, they expire and are discarded. The problem is that high bandwidth applications, especially peer-to-peer (P2P) applications can often shove hundreds of large 1472-byte packets into the pipeline at around the same time. Because routers and switches can only transmit one packet at a time and basic routers only handle packets in the order which they come, VoIP and online gaming packets can be delayed for an additional 10 to 1000 milliseconds which causes severe degradation. This phenomenon is called jitter and it is best explained in this 7 minute animation on what jitter is and how it can be solved. Without network prioritization, the fairness of bandwidth allocation is dictated by law of the jungle where big bandwidth applications starve the small bandwidth applications.
The fairness of prioritizing low bandwidth applications.
It is unfortunate that some Net Neutrality advocates can’t even acknowledge the fact that low bandwidth applications need protection from high bandwidth applications. Yet their concerns are based on misinformation rather than good engineering or fairness. There should be no question that it is always fair to protect low bandwidth applications with expedited packet forwarding, yet Net Neutrality advocates question this practice.
We can draw a real-world analogy to explain the fairness of the situation. Imagine if we had two roommates Victor and Peter sharing a single phone line and paying an equal share. Victor represents VoIP and Peter represents P2P. Victor occasionally needs to make a 1 minute phone call every 25 minutes to conduct his business, but Peter likes to frequently leave the phone running continuously for 100 to 500 minutes at a time all day and all night even though he’s not actually attending the phone. In order for Victor to squeeze in a phone call, he is forced to wait for Peter to hang up his phone before he can use his own phone to get a single call in, so he is frequently forced to give up using the phone as many as 20 times while Peter leaves his phone running for 500 minutes even though Victor only needs to use the phone 4% of the time.
Peter would not be negatively impacted if he had to give up using the phone one minute out of every 25 minutes since his phone is running in the background, but there’s no mechanism in place to kick Peter off the phone line. So the network operator steps in and says that it might be worth implementing a system that allows Victor to jump in for a minute over Peter’s background phone call and automatically resume Peter’s phone connection after one minute when Victor hangs up. The network operator could in theory just let Victor suffer which might spur Victor to purchase a new separate phone line that doesn’t suffer from Peter’s network hogging, but it decides that good customer service is more important in the long run.
Just when things begin to look up for Victor, political activists rushes in and declare that it would somehow be unfair to to give Victor priority over Peter and that that the problem should be left between Victor and Peter to solve even though Victor has never been able to overcome Peter in the past. These activists want the government to craft rules prohibiting the network operator from protecting Victor and they even argue that Victor should simply accept the situation. The analogy of Victor and Peter mirrors the situation where political activists argue that regulations are needed to prohibit network operators from protecting VoIP from P2P applications because they argue that VoIP can protect itself and that network operators shouldn’t discriminate against P2P.
Misleading arguments against prioritization
Many Net Neutrality advocates who are not engineers often repeat the myth that VoIP applications can be engineered to tolerate jitter. While it’s true that a mechanism called jitter buffering can reduce the likelihood that packets are discarded, use of the jitter buffer increases the baseline latency which results in delaying the entire conversation. So while in theory it is possible that we can set the jitter buffer to 500 milliseconds (200 is the practical limit) so that all packets that are delayed less than 500 milliseconds aren’t discarded, we end up delaying the entire phone conversation by half a second which is a horrible way to hold a phone conversation. Games are similar in that they can also be engineered with jitter buffers to reduce packet discards, but the game quickly becomes unplayable as the baseline latency rises.
Another example of misinformation was when Free Press lawyer Marvin Ammori on February 24, 2009 at a Federal Communications Bar Association (FCBA) panel argued that Vonage doesn’t need QoS for their VoIP application. But this is false because Vonage actually prioritizes their own VoIP traffic over other Internet traffic using the Analog Telephony Adapter (ATA) they issue to their customers. This ATA acts as the broadband gateway to the Internet for millions of Vonage customers so Vonage is in fact favoring their own application over others because it’s fair and because they have the power to do so.
Others argue that P2P applications like BitTorrent have been reengineered to be “network friendly” so the network operator don’t need to do anything, but actual test data and shows that the new BitTorrent protocol is just as unfriendly as before. BitTorrent the company even acknowledges that their new “network friendly” protocol has little effect on upstream jitter and even less effect on downstream jitter, and they acknowledge that they haven’t done anything about BitTorrent’s tendency hog 90% of the bandwidth at the expense of other applications like Web Browsers.
The fear of network operator abuse is without merit
Some argue that if we allow network operators to favor some applications over others, they could abuse the situation for anticompetitive reasons. But this argument falls apart for several reasons.
- Network operators have no incentive other than customer satisfaction to protect VoIP and gaming applications. If anything, their incentive is to do nothing because that would mean their own circuit switched phone networks look much better than VoIP service competitors. That’s because Telco and cable operators run their analog or digital phone service on separate dedicated frequencies that aren’t negatively impacted by congestion on the broadband service.
- If a network operator misclassifies low bandwidth applications like VoIP as a low-priority application, this would be an obvious case of anti-competitive behavior and the FCC could step in and address the problem.
- The FCC could have an anti-discriminatory rule but spell out but it should have an exception under “reasonable network management” that it is reasonable to favor low bandwidth applications over high bandwidth applications, and that it is reasonable to try and balance out the amount of bandwidth consumed by every application. The FCC has already allowed Comcast’s new “Fair Share” network management system to prioritize low bandwidth subscribers over high bandwidth subscribers, so the same concept of fairness should apply to applications as well.
- Users and Application developers could still be given preference to set their own priority preferences with the use of DiffServ (DS) or Type of Service (TOS) fields so long as they operate within their allotted priority budgets. This gives ultimate control to users and applications to override any ISP behavior so the concern that ISPs might abuse the system is without merit.
- Just because it’s possible for users and applications to set their own priority levels doesn’t mean that there is no role for the ISP to play. We know that the majority of applications and users don’t know how or don’t care to set their own DS or TOS fields, and the network operator is the only entity that can deal with these users and applications.
We are at a critical juncture where the FCC has a major opportunity to set some sensible rules of the road. But to be sensible, the rules must be based on good data and it must allow for reasonable network management where the definition of “reasonable” is crucial. The problem is that many advocacy groups have put forth differing sets of “data” and “facts” so the challenge is to determine which facts are true and which are myths. This article dispels many of those myths and we at Digital Society hopes that the FCC will allow the Internet to be hospitable to everyone and every application, and not just the ones that are the most aggressive while falsely claiming to be “network friendly”.