IP Convergence: Beyond VoIP, Beyond Cost Savings

I was reading some Ronald Reagan quotes that a family member forwarded me recently, and as I read the following quote I couldn't help but think of one of the hottest topics in the blogosphere right now, net neutrality.

I haven't really weighed in on net neutrality in my own blog yet, even though I've posted comments in a few other places such as the Pulver blog, Susan Crawford as well as chiming in on the Global Crossing Podcast. Like any large complex issue, I think it's easier to discuss some of the finer points of an issue than it is to try to discuss all aspects of the issue at once.

I hate to dissappoint, but I'm not going to attempt to do that at this time either, or I fear I might overload the input buffer on the web server after hitting the submit button (I've already lost this blog post once today when my computer reset, argh!) I will say that I tend to agree with my blogging colleagues on this matter, Paul Kouroupas, Onofrio "Norm" Schillaci, and Adam Uzelac: no matter what the final solution may be, I believe that legislating/regulating it is a huge mistake.

What I want to talk about in this post, however, is what these tiers of the Internet might look like, and what the business issues are in an absense of regulation.

When people talk about tiers in the Internet in the context of net neutrality, they are talking about what engineers refer to as QoS, and in particular, a method of implementing QoS called Differentiated Services, or Diff-Serv. Each IP packet is marked, or painted, in such a way that it is identified as being part of a Class of Service. There are 3 bits available in this field of the IP header, so you can theoretically have up to 8 Classes of Service. When a router in the service providers core receives a packet with the marking, it figures out which output interface it needs to go to and copies it into one of the multiple queues that have been defined for each type of service. The queues are then drained according to an special sauce algorythm that insures that the highest priority queue gets any attention it needs while the lowest queue takes whatever is left after the other qeuue(s) have taken care of business.

This is the technical capability that would enable the likes of AT&T and Verizon to implement their scheme of priorities, but if you think they are waiting to implement it until after they've won this net neutrality debate, you are probably wrong. We have already implemented this at Global Crossing, and have had it working for many years now in fact. We have 3 Classes of Service, the highest priority being for our VoIP Platform (which also carries over 50% of our ordinary long distance traffic), the lowest priority for Internet Services, and our VPN customers get to buy out both of those classes (Premium and Basic, respectively) as well as one that resides in between the two known as Enhanced.

Global Crossing continues to have one of the highest quality Internet backbone in spite of it being in the lowest priority class, because we continue to design the network to perform at a high level of quality. So why even have Classes at all if the lowest one isn't degraded? Well, it does degrade every now and then, during severe fiber cuts for example. It isn't often, but when it happens, you're glad those voice calls are still working.

(for the sake of the hypothetical scenarios posed below, I will use ATT, Global Crossing, and an unnamed content provider for illustrative purposes).

The issue of peering is this. If ATT demands that content providers pay them in order to put them into the premium class (similar to how our IP-VPN customers pay us to put traffic in the premium queue, but on a whole different level, then that content provider has to connect directly to the ATT core (i.e. become a directly customer of theirs). Unless QoS/CoS is enabled across the existing peering connection that exists between ATT and Global Crossing, and then a content provider could connect to Global Crossing and we could transport the packets between the content provider and ATT at a premium level and give them to ATT at a premium level.

The kicker is that today peering amongst the top providers (known as the Tier 1, not to be confused with the other use of tier in this post) is settlement free. So here is where the business issue enters the equation. If the Best Effort Internet traffic remains settlement free, what is the charge for the premium traffic, and how will it be billed? By the bit? In both directions? Do we bill ATT for the premium traffic that we receive from them and vice-versa? But what about VoIP where the traffic is roughly equal bi-directionally? How does that help ATT if the billing is a wash? Believe it or not, these are the same discussions that have raged in the Internet peering debate since the Internet went fully commercial in 1995. Ultimately, we all decided that peering should remain settlement free, although some providers do requirement settlements if the ratio of in to out traffic is too high because of the unfavorable economics.

What if ATT billed the content provider directly and Global Crossing was just a pass-through, with a contract with ATT to ensure that we didn't give them any premium traffic that wasn't being billed. Then there are the billing mechanics to content with...ATT has to know which content provider to bill for what amount of traffic, since they will probably get multiple content providers sending them premium traffic over the same link.

There is also the technical issue of insuring that the content provider that pays for the priority ride on ATT's network gets it in both directions. In order for that to happen, ATT has to identify the content originating at each broadband connection and identify it as premium traffic. This is not so easy in current edge equipment...it is not only processor intensive but it requires a lot of administrative overhead. The lists of IP addresses that identify content providers and their applications will need to be updated across all customer connections every time a provider connects or disconnects from this premium service.

Okay, for now let's assume that we can get that all sorted out so that ATT can collect additional revenue, either by billing us and Global Crossing passing that cost on to the content provider, or by collecting it directly from the content provider.

How does the content provider who is either still trying to turn a profit or turning a profit (but not an unreasonable one) going to pay this added business expense. They probably can't absorb it in their current business model, and even if they can to maintain margins they will have to raise prices. Since most content revenues are derived from advertising (see Andrew Odlyzko's paper Content is Not King), they will have to raise their ad rates. I'm not sure if advertising will pay more, my understanding is that it's a pretty competitive business as it is.

While other technologies like ye old Infranet iniative or IMS may have something to say about improving the billing mechanics around this, the business implications are still something to be concerned about.

I will leave you with one further quote:

Don't let them exclude the VON Coalition from the discussion Jeff, throw some Reagan at them!