Why Depeering Matters
According to a couple of people who wrote to me, depeering is a big problem that will get bigger as the two mergers (MCI-Verizon and AT&T-SBC) roll on.
The key issue is cost, not (necessarily) content control, although content/application control may be a byproduct. Depeering will allow the two mega peers to make it very expensive for independent actors to reach the mega peers' subscribers. Both Verizon and SBC have voice and video services they want to protect, so it's in their interest to make it expensive for new large voice or video providers to reach their huge numbers of users.
Briefly, backbones connect either through either peering or transit arrangements. If they're peering, backbones agree to exchange traffic with each other (only each other's traffic) at no cost. If they have a transit arrangement, one network pays another for the right to connect. The transit provider then makes sure that the paying network's packets get to their destinations.
If the two mega-peers depeer, then other carriers will have to pay about twice as much to have other backbone providers carry their traffic. This will give all the carriers involved lots of incentives not to allow unauthorized bandwidth hogs (like competing video services) to ride across their networks.
Peering was always a matter of private contractual arrangements, so it's not clear what hook a regulatory agency would have to second-guess a mega peer. (There are no domestic or international regulations that govern how backbone providers interconnect.) You'd have to argue that the mega peer was acting anticompetitively, using its market dominance to extract high prices and drive its competitors out of business. But both MCI-Verizon and SBC-AT&T will argue they don't dominate the market for backbone access (although they may have the lion's share of internet subscribers).
And the mega peers can assert that they're changing the world by making the internet a “quality of service” network. This is the new raging meme. The internet never promised that your packets would get where they were going. Now backbone providers can promise new services (video, voice, data) only to their subscribers, and say they're going to work blindingly fast. Competing backbones won't necessarily provide the same services. This will fracture the internet, if the backbone provider can find ways to survive without interconnection (or by only connecting to networks that agree to the backbone provider's policies).
In the absence of dominant backbones, none of this is a problem. Markets will find ways to satisfy consumers, and users can believe that they're able to “see” the entire internet. It used to be in the interest of backbones to have many many peering relationships to keep their customers happy. But if one or two dominant backbones emerge, they won't feel the same pressure of externalities.
There are now only five Tier 1 (highest level) backbone providers (this is from a 2003 article by Michael Kende): Cable & Wireless, WorldCom, Sprint, AT&T, and Genuity. If the large ones get so large and powerful that they no longer feel the need to interconnect (and can charge high prices for their new services), they can depeer with all smaller backbones, exact high prices for transit, degrade the quality of their interconnection with the smaller backbones, or take any number of other anticompetitive steps to protect their private prerogatives.
So that's why depeering matters.
Should we fear mega peers, or just laugh it off?
In a Sept. 14 filing with the FCC, a company called Broadwing [warning, high Flash content] said the following:
We are concerned that the mergers of two significant Internet backbone providers (AT&T and MCI) with two of the largest providers of high-speed Internet access (SBC and Verizon) will result in two “mega peers” with the potential to de-peer competitors such as Broadwing and dominate the market.
Broadwing operates a fiber-optic network connecting U.S. cities, and offers services (VPNs, managed network services) to a variety of customers.
Verizon responded on Sept. 21, saying that the “transaction will not give the combined Verizon/MCI market power in the Internet backbone business. The combined company will carry less than 10% of North American Internet traffic, it will rank fourth among seven comparable or larger backbone operators, and operators other than those seven will carry approximately 35 percent of Internet traffic.”
What's the significance of a “mega peer”? Wikipedia tells us that peering arrangements are private contractual arrangements that exist between peers that have “roughly balanced” traffic. One peer doesn't charge another for carrying the other peer's traffic and handing it off to yet a third network.
What Broadwing is worried about is that extraordinarily large amounts of backbone will be controlled by two players, who will then decide to “de-peer” — disconnect their networks — from other networks whose favor or traffic they don't feel they need. The result for Broadwing will be that life gets more expensive. They'll have to pay the backbone provider — just like small ISPs have to pay upstream providers — to carry their traffic.
In the words of the Wikipedia hive mind:
De-peering usually results in degraded performance for customers as traffic is forced to take less optimal paths; in the worst case some parts of the internet can become unreachable to hosts on either side of the de-peered networks. This occurred in October 2005 when two backbone NSP's in the USA, Level 3 Communications and Cogent Communications, de-peered following a commercial determination by Level 3 that it was significantly larger and therefore Cogent should pay it to continue peering.
Another fear might be that sufficiently powerful backbone providers (especially those vertically integrated with major broadband providers) could just decide not to connect to the rest of the internet. Verizon/MCI could decide to connect only to AT&T/SBC, and to create a gigantic (but limited) U.S. walled garden for their own proprietary services (no Skype; only approved video — all the Barton-Dingell steps).
But this is absurd, isn't it? Wouldn't it be a terrible business decision for a backbone provider to cut off traffic? Why would any ISP want to do business with a backbone that doesn't actually have arrangements that allow it to ensure the carriage of packets around the world? Don't we have a flat world full of global transactions?
If email doesn't get where it's supposed to go, then people will find other ISPs and (indirectly) other backbone providers.
Laughable, right? There will always be other ways to go to get on the backbone network. If a “mega peer” tries to take advantage of its position to finally put a chokehold on the network of networks formerly known as the internet, people won't stand for it. They'll find other ways to get their packets where they're going. The thing about peering is that there are always other players to peer with.
Back me up here — isn't there sufficient competition in the backbone market? Won't it be impossible for any US “mega peer” to make our experience of ”the internet” into a single huge walled garden?