Carrier Ethernet

From the carriers’ point of view, it was a tale of two technologies: It was the best of inventions; it was the worst of inventions. It drastically reduced some costs, but threatened an established business model. Fiber optics certainly did spark a revolution in telecommunications, but counter-revolutionary forces of monopoly have been doing their best for the past 30 years to keep it at bay and preserve the old way of doing business. 

And while they have not been entirely successful, their efforts, and the regulators who protect them, are still impeding progress. The incumbent carriers are using pre-fiber pricing models, based on voice-equivalent channels, to gouge customers of their monopoly facilities, especially when Special Access rates apply.

Fiber optics were a revolutionary technology, like the integrated circuits that computers are made of. But we have only seen part of the fiber dividend that should have resulted from their deployment. We’re paying copper prices in a fiber world. While available fiber speeds have risen, carrier services haven’t keep up with Moore’s Law, especially in the United States. And it’s largely a national phenomenon: While the US once led the world in “broadband” services, and pioneered competitive telecom, domestic services are now costlier and slower than those in most other industrialized countries. What we see instead is a growing digital divide between the higher speeds and lower prices available on fully-competitive routes, where Moore’s Law is allowed to operate, and the atrophy of the ILEC monopolies’ ancient regime. 

Living in a vacuum-tube past

In the days before fiber optics, transmission was by far the most expensive component of the telecommunications industry. A long-distance phone call was expensive to deliver because the transmission facilities were themselves costly. Before the 1920s, every call required separate wires. Then multiplexing was developed. The analog multiplexing systems of the mid-20th century treated phone calls like radio channels, though the capacity of most coaxial cable and microwave radio transmission systems was quite limited. As the cost of transmission gradually fell, so did the cost of long-distance calling and leased circuits, as tubes gave way to transistors and digital systems like T1 began to appear. 

But some of the savings were not distributed in the form of lower costs. While the capacity of transmission systems was growing even before fiber optics arrived, the labor-intensive cost of stringing wire to the home continued to rise. In order to hold down the basic monthly telephone rate, regulators distorted the Bell System’s cost allocations. Long distance subsidized local more and more of the local bill. And anything that reduced the volume of long distance calling, like a leased line, was also expected to make a contribution to make up for the calls it could be displacing.

That might have made sense in the 1980s, when voice was still the dominant use of the network, but it’s ridiculous nowadays. Yet that is still the basis of today’s “special access” tariffs. These are the leased line rates charged by local telephone companies for either a regional circuit or their end of a long-haul circuit.   Long distance calls pay an above-cost fee called switched access; special access is the rate charged for circuits (“special services”, in old telco-speak) that “bypass” it. The fact that data circuits and today’s high-speed Internet use thousands of voice calls’ worth of bits, but it doesn’t cost that much more to provision, doesn’t factor in much. It’s all about the call, preserving the lost revenue from unmade toll calls. 

Toll bypass is only a small fraction of special access is used for nowadays, especially since switched access rates and thus tolls have come down. Special access is mostly used by cellular phone companies to connect most of their towers to their own networks, and by ISPs to reach their commercial customers. 

Lighting the fiber has gotten much cheaper

Fiber optic technology has advanced rapidly since first going commercial in the early 1980s. The first systems installed by the Bell companies typically carried 135 Mbps, some maybe twice that. In the early 1990s, SONET came on the market, with high-end systems running at 2488 Mbps, a speed known as OC-48. This could carry 32,256 telephone calls using standard multiplexing. This seemed almost unlimited at the time. But today it’s obsolete. High-speed data systems run much faster.

Fiber optics don’t have infinite capacity. It’s just huge. What’s nice is that the cost of the optronics, the stuff needed to light it, is almost falling along Moore’s Law lines. They’re made from semiconductors, after all. So the number of bits you can cram down a fiber is growing as the price per bit falls. Not quite as fast as computer memory, because real optics are involved, but still pretty fast. Take SONET gear. You can still find these old rates in the AT&T (News - Alert)-Texas tariff. In the mid-1990s, they set the price to install a first-generation OC-12 mux in a virtual collocation arrangement at over $60,000 up front plus a rental fee of over $2000/month. Perhaps some suckers are still paying that rate – I just found it on a current bill, for a network a client was acquiring. Today a better-quality SONET mux can be bought new for around $6000. Add a few grand to raise it to OC-48.

10G Ethernet is priced at roughly what 100 Mbps cost in the 1990s. And that’s now considered low end. Dense Wavelength Division Multiplexing (DWDM) is where the high-end action is. This typically fits 40 separate wavelengths (lambdas) onto a single strand, each capable of 40 or even 100 Gbps, if you spend the money. To be sure, the 100G cards are still exotic and rare, while 40G cards are not yet common. Each such card – one is needed at each end of the connection, though it can pass transparently over many intermediate nodes and amplifiers – will set you back as much as a nice car. (Or a month’s rent on an OC-12 crosstown ring from AT&T.) But equipment prices are falling rapidly. If you put 40 lambdas of 40 Gbps onto a strand, it’s carrying 1.6 terabits per second. Depreciated over a few years, it’s not much per bit.  This is what fuels the growth of the Internet backbone.

Applications move to on-fiber data centers

So the transmission services market is now divided into two categories, competitive routes and monopoly routes. On the competitive side, such as between points of presence in major cities, prices are usually low. This even applies internationally. So a circuit from a major building in New York to London or Chicago or Miami is pretty cheap. As a result, a big business in collocation sites – data centers and carrier hotels – has sprung up at buildings with competitive fiber access, usually in or near major cities. Web hosting and other high-volume network applications almost always take place in these sites, not at their owners’ offices.

But if you have to reach an actual customer location, a building not on a major fiber route, all bets are off. These monopoly routes are often priced at special access rates, so a DS-3 (45 Mbps) circuit typically costs around $3000 plus $100/mile per month. Just getting from one side of the city to another can cost more than the competitive route across the country. Carrier Ethernet, totally unregulated, is priced all over the map. ILEC list prices are high (akin to special access, thousands of dollars for month for a crosstown connection) but discounts of 80-90 percent are common along the most competitive routes.

Pricing is based on tariffs of the distant past

Why are these rates allowed to be so high? Until the early 1990s, the Bell companies were on rate-of-return regulation, so the return on their investment in any given “basket” of services was supposed to approximate a target. If they invested more, they made more profit. They then convinced the FCC (News - Alert) and most states to allow them to adopt price cap regulation, in which their prices would adjust annually based on inflation minus a productivity factor. So their capped prices have been based on 1992 norms -- before there even was a public Internet, and when the really fast backbone links ran at only 1.5 Mbps. There were fewer than 20,000 cell sites then, vs. over a quarter million today. Even SONET had not yet been rolled out nationwide, and early voice-oriented fiber terminals were the norm. Special access was indeed largely a toll bypass service. So rates that might have given an 11-12 percent rate of return then translate today to a rate of return of over 100 percent, given the vastly higher volumes of special access traffic and the lower cost of lighting the fiber. That’s monopoly at work. Not that the Bells and other big ILECs are even required to do the accounting any more.

The irony is that the Bells are whining to regulators about how they’re losing money hand over fist. This is a remarkable feat of creative accounting. They’ve left the costly old plant on the books of their state subsidiaries, but moved most of the revenues out. Since most high-speed services are now deregulated, they’re not counted as “telephone company” revenues. Only dial tone and a few state-jurisdiction access services are. The preponderance of data circuits is jurisdictionally interstate, though. Hence the state-regulated entities pretend to be losing money while the parent companies are reporting record profits to Wall Street.

And if federal price caps seem generous, bear in mind that in most metropolitan areas, the incumbents have been granted “relief”, which allows them to set prices wherever they want. They first claimed that this would let them respond to competition and price under the caps, but in most cases prices are well above the caps. That’s why they can charge over $200/month for a DS1 (1.544 Mbps) local loop, which is really just a type of DSL circuit, using the same type of wire as a voice line.

If all basic transport were really competitive, then prices would naturally fall towards cost, as it has done on the major long-haul routes. But instead the divide grows greater by the year, demonstrating how competition has failed in most local markets. Bells milk more and more monopoly profits, while long-haul carriers scramble for a share of a competitive market. So Level 3, the biggest domestic backbone player not affiliated with a Bell company, has never made a profit, though they’ve managed along the way to roll up the fiber assets of several failed competitors. But the Bells have bought up the assets of the two largest long distance companies (Verizon’s (News - Alert) purchase of MCI and SBC’s of AT&T) and leveraged their control over special access to cause T-Mobile to pull out of the country, while Sprint barely hangs on.

Stimulus awards provide unequal help

The $7 billion broadband stimulus subsidies, awarded in 2009 and 2010, provide for the construction of a number of new network facilities. Much of this is “middle mile” fiber, which precisely targets the mileage portions of special access, and some goes to “Community Anchor Institutions”, large users whose last-mile connections are counted as middle mile for grant purposes. This provides the first competition to the ILECs in some areas and may even be the first high-speed facilities available at all in some rural areas where ILEC facilities are particularly thin.

So will these bridge the divide to their users? Alas, not always. These projects have had to file “sustainability” plans, basically business plans with proposed rates. Stimulus projects can be put into two categories. Some are truly disruptive, offering dark fiber or high-speed service at a low cost. Maine Fiber Company’s build, for instance, will pull 288 strands of dark fiber around the state, allowing its customers to light the strands however they please. But many of the projects are simply using “Bell minus” pricing, with special access tariffs as a benchmark. A 15 percent discount off of a ridiculous price doesn’t change much; it may take away much of the existing business, but the users are still at the mercy of historical long distance rates. 

So long as the Bells are allowed to price their local facilities as if fiber optics had never been invented, and potential competitors treat them as a price umbrella whose rate structure should not be disrupted, service in the US will continue to suffer, and both competitors and consumers will overpay.