Postmodern network architecture and the failure of reregulation 

Fred Goldstein,  April 2003


... Most wire centers predate 1950; the tradeoff was based on parameters of a long-gone era.  


...a truly forward-looking postmodern network doesn't need nearly as many wire centers as the existing "modern" network has. It just needs a lot of DLCs, fed from fewer, larger regional wire centers....

I am particularly fond of addressing topics that involve a combination of technical, business and regulatory factors.  While it is often difficult to separate these whenever dealing with telecommunications, few topics intertwine the three of them as closely as network architecture, the fundamental standards and assumptions by which networks are designed and built.  The AT&T divestiture of 1984 forced a rethinking of network architecture, at least at the interexchange level. The Telecom Act of 1996 changed the regulatory treatment of local networks, forcing changes in the way networks are interconnected.  But it did not seriously change the architecture of local networks themselves, nor did it anticipate such changes.

Indeed one of its potential flaws, or at least areas where regulatory interpretations of the Act have come up short, is in the way it simply assumes, or in some cases locks in, obsolete architectural concepts.  This is no doubt not caused by malice; more likely, the lawyers who draft regulations are technically savvy enough to understand at a high level how actual networks are built; they are not however always well versed on why networks have been built a certain way, or why those assumptions may have become obsolete.

This leaves us with the situation that we now find ourselves facing, in which the FCC has, for all intents and purposes, preserved competition in the embedded, aging local network while giving ILECs a virtual monopoly in newly constructed networks.  And in so doing, regulations still talk about "forward-looking" costs that are more accurately described as based on the state of the art ca. 1996.  

Such "modern" (they're downright bleeding edge, by some ILEC standards)  networks are based on a concept called "Carrier Serving Area" (CSA), a concept from the 1980s designed to support ISDN while minimizing the cost of the local network's outside plant.  A CSA network design limits the length of local loops to 12 kilofeet.  A traditional telephone switch goes into a central office building (wire center), with copper loops serving subscribers within 12 kf. Beyond that, loops are served out of Digital Loop Carrier (DLC) systems, which are fed by optical fiber. SONET rings connect wire centers to each other.  ADSL was invented in the early 1990s with CSA in mind; monopoly LECs were expected to deploy DSLAMs in the remote terminals, feed them with optical fiber, and use them to deliver video.

Before DLC took off in the 1980s, local loops had to connect directly to central office switches.  So wire centers were laid out to trade off the cost between more switches vs. longer wires.  Most wire centers predate 1950; the tradeoff was based on cost parameters of a long-gone era. The FCC's TELRIC wholesale-cost methodology for selling unbundled network elements to CLECs is based on "green field" costs of a whole new network, built according to CSA, but one that preserves all existing wire centers, which were laid out before CSA.

The wire centers themselves, the big buildings with the telephone company's name on the side, are now obsolete.  Most wire centers were built to house large electromechanical switches; the oldest were built to house rooms full of operators sitting at cord boards.  Modern telephone switches are different.  They're compact -- 50,000 lines can be switched in a single seven-foot rack!  They do this because they separate the physically-bulky line terminal (essentially a multiplexor for analog lines) from the high-capacity exchange terminal, an all-digital system that terminates only high-capacity digital lines.  A DLC is the exchange terminal in a CSA network, so long as it is installed using the preferred, less costly integrated method, which digitally connects it to the switch, with no analog interconnection.  

So what function does a wire center now serve?  A switch has to go somewhere, but with dirt-cheap SONET bandwidth between virtually all wire centers, there's no need to actually put a switch in a wire center unless, as in core urban areas, a large number of lines physically terminate there. In smaller wire centers, the "switch" nowadays is usually just a remote node off of a large one anyway.  For the smaller sizes, it can even be just one or more DLCs.  (The Telcordia GR-303 standard for Integrated DLC supports up to 2048 lines fed by up to 28 DS1s.)  That's essentially how CLECs are required to access the local loop.  Thus the wire center is just a big DLC hut, literally for CLECs, figuratively for most ILECs.  And so a truly forward-looking postmodern network doesn't need nearly as many wire centers as the existing "modern" network has. It just needs a lot of DLCs, fed from fewer, larger regional wire centers.  Many suburban and most rural wire centers are thus obsolete. Regional switches can serve DLCs spread over a large geographic area.  Even if there are multiple rate centers nearby (another artifact of bygone days), one modern switch can serve many prefix codes.

So far this all sounds good; such progress lowers the cost of networks.  And that's good for the telecom business; lower cost is an incentive to invest in network modernization.  But now the regulatory problem comes in.  CLECs get their best access to local loops at the old-fashioned wire center. ILECs have to permit CLEC collocation there; if the ILEC claims there is no space, they essentially have to jump through hoops to prove that none can be found.  Under the FCC's latest ruling, CLECs lose DSL access to "hybrid" lines -- those behind fiber-fed DLC. They can probably still get voice-grade POTS access, at least if the subscriber site was in place before 2003, but POTS by itself is a tough business for CLECs.  It is theoretically still possible for CLECs to rent DSL-qualified copper subloops behind DLCs, but since DLCs are usually in pedestals, manholes (er, "controlled environement vaults") or at best, small huts, collocating there is often difficult or impossible.  But these remote terminals are today's real wire centers. And a lot more are coming; the FCC has just given the green light to ILECs to build more of these in order to gain a competitive edge against CLECs.

That's the conundum facing postmodern networks.  Progress depends on competition -- ILECs have amply demonstrated that absent competition, they much prefer to live in the past.  But the obvious architectural direction, moving to a postmodern architecture with fewer wire centers, perversely penalizes competitors.  The FCC views this as an incentive for ILECs to move forward -- if ILECs take one step to modernize now, they can rid themselves of pesky CLEC competition once and for all.  Of course there will then be little incentive to take advantage of other gains that might come from postmodern architecture.  The FCC's rules don't recognize the obsolesence of the existing network; its new direction is anticompetitive and thus slow progress, harming competitors and consumers at the same time.

A good fix would be to separate the outside plant from the core network. I've already suggested here how such a scheme could work. A simpler fix is to allow "virtual" usage of remote terminals by CLECs. While ILECs claim that unbundled network element prices are below cost, that's hard to argue when these are set on "forward looking" costs, which, for new construction, are current costs. The issue isn't technical, it's political.