Since the inception of the phone system, companies like AT&T have built large central offices (COs) to house huge boxes full of switches. These switches were designed to intelligently create a circuit between a person dialing a phone number and the person whose location the number represents.

A large “switch” (as they’re commonly referred to in the telecom industry) like the ESS Number 5 consumes 1,000 square feet or more of floor space, costs several million dollars, and can handle 50,000 simultaneous phone conversations, or active circuits.

Like the minicomputer and mainframe before them, these circuit-switching mechanisms will soon be facing stiff competition from systems based on microprocessor technology and software-based switching technology.

Circuit switching versus packet switching
Traditional circuit-switching telephone technology attempts to take your voice and create a complete circuit between the originating phone and the destination phone. There’s no conversion of the voice from analog to digital as part of the system.

The technology has been stable and reliable for decades, but with the explosion of cell phones, pagers, and other devices requiring new phone numbers, the existing switch networks are being severely taxed. Given the cost and real estate requirements to build out additional switching capabilities, the industry has been looking to Voice over Internet Protocol (VoIP) technology to allow for rapid expansion of capacity and instantaneous reconfiguration of the network using software.

VoIP uses “packet-switching” technology to move voice conversations between two end points. The voice conversation is first digitized and then placed into a series of TCP/IP packets that can be routed across a data network. But unlike standard data packet routing, where delays can be tolerated, voice packets must be given priority when reaching their final destination or the voice will be broken up and unintelligible.

Software like Microsoft NetMeeting uses the same basic technology to allow two people to talk over the Internet today. But this is a simple point-to-point connection that’s trivial compared with the engineering required to replace a massive circuit-switched network with a packet-switched one. There’s a new breed of company that I call an NG-CLEC (next-generation competitive local exchange carrier) that’s attempting to do this today.

Implementing the new technology
NG-CLECs are implementing their packet-switched networks by putting in a new CO with packet-switching technology that can handle the same 50,000 simultaneous phone conversations as the circuit switch mentioned above, but for less than $1 million—and it’s the size of a four-drawer file cabinet.

The NG-CLEC connects small businesses to the new CO (which is now a data center instead of a traditional telco point of presence) using proven and reliable T1 connections. (Companies implementing these services aren’t relying on new technology like DSL or wireless connections yet—neither has proven reliable enough support a high-availability application like voice traffic.)

A standard T1 supports 24 x 64 KB channels, or about 1.5 megabytes of transfer speed. In order to digitize and carry uncompressed voice traffic, each voice conversation requires about 64 Kbps of bandwidth.

NG-CLECs are using client-side routers that have the technology to digitize voice streams and, more importantly, use the remaining available bandwidth for Internet connectivity. The NG-CLEC configures the client-side router to allow a combination of available phone lines versus Internet access speed. For example, if you configure 12 phone lines, you’ll have about 750K of available bandwidth remaining for Internet connectivity.

The client-side router can dynamically allocate bandwidth and assign it to voice channels as required so that if no one’s using the phone, you can get the full 1.5MB available for Internet connectivity. The routers break out analog lines that can be connected directly to a local PBX for integration into the local phone network. I expect to see these companies add new routers that double as PBXs or to start selling PC-based PBXs that integrate with their offerings directly.

As an NG-CLEC builds out its network into multiple markets, it can push long distance traffic across its own network backbones. Where it doesn’t have a presence, it simply attaches to the existing circuit switched network to pass the long distance traffic.

As its network grows, its cost of providing long distance traffic decreases and its profitability increases. So there’s a huge push by an NG-CLEC to get as much coverage in as many markets as quickly as possible.

Saving money with NG-CLEC services
In 2001, several companies are rolling out these services in major metropolitan markets. The offerings look strikingly similar regardless of the company.

By using a combination of new packet-switching technology and proven, reliable T1 client connections, they’re able to provide a package of services for a fraction of the normal cost.

For example, standard business-class telephone lines average $50 per month, and a 512K DSL connection averages $350 per month. For an office that needs eight business lines and a 512K Internet connection, this would cost about $750 per month. NG-CLECs, however, are offering the similar functionality (eight phone lines, 1MB Internet connection) for about $450 per month. Because of their ability to pass long distance traffic (in the markets they serve) for no marginal cost, the NG-CLECs are also offering 1,000 to 3,000 minutes of long distance at no charge with additional minutes at five cents or less!

This technology is a godsend for small businesses. They can get high-speed Internet access and local phone service for unbelievable prices. For large companies with remote offices, this can be a very low entry cost.

It’s also technically possible for the NG-CLECs to link your remote offices to your central office in such a way that the extensions in the remote offices appear to be local. None of the existing NG-CLECs are offering these services now because they’re focused squarely in the 10- to 50-person office space. But by using the same technology, they can drop one or more T1s into a corporation’s telecom cabinet and route traffic between the main office and remote offices. I believe this will be the first major implementation of VoIP across wide area data networks because—unlike the Internet—this data/voice traffic can be easily prioritized.
What type of connection have you chosen? DSL? Cable? T-1? Was this your first choice? Are you thinking of changing it? Tell us in an e-mail or start a discussion below.