Intelligence at the Edge #6


CANARIE propses gigabit Internet to the home while U.S. telcos diddle with DSL.

David's smiling face

By David S. Isenberg                                                              amnetlogo

From America's Network, February 1, 1999

When Bill St. Arnaud tries to show earnest telco types the leading edge, he might as well be talking Martian. When he explains how he’ll deliver gigabits via fiber to the home (FTTH) for about the same cost as megabits via digital subscriber line (DSL) or cable modem, their minds seem to be stuck in the traffic jam at the intersection of IP and SS7.

In the midst of the distracting pseudo-battle between DSL and cable modems, it’s hard to remember that FTTH is still the broadband endgame. Despite the pall of failure around early-1990s interactive TV, the supremacy of fiber has been clear as glass for over a decade.

St. Arnaud, the mild-mannered director of network projects for the Canadian Network for the Advancement of Research, Industry and Education (CANARIE; has not lost sight of the FTTH beacon. The newest CANARIE project, CA*Net3, will throw away synchronous optical network (Sonet) and asynchronous transfer mode (ATM) to become the world’s first all-optical Internet backbone. St. Arnaud believes this design can be extended into the home. He proposes to throw away DSL and cable modems, too, bringing CA*Net 3’s all-optical multigigabit Internet into every Canadian home by 2005.

WDM in local networks In long-haul networks, wavelength division multiplexing (WDM) has increased fiber capacity by a couple of orders of magnitude in two short years. This year, a single fiber will have throughput for 15 million calls — enough to handle the entire U.S. busy hour.

But WDM has not yet hit the local loop. St. Arnaud thinks it’s because established providers are tangled in reuse of their own nets. Cable companies have cable modems, so data service can run on existing, broadcast-oriented networks. Telcos have DSL, which is backward-compatible with twisted pair. For both, the key word is backward.

The same goes for transmission protocols. Sonet was designed for reliable voice (connection-oriented) networks. ATM’s goal was a single protocol for handling voice, video and packet services. Neither anticipated Internet Protocol (IP).

Sonet and ATM become shaky when they’re not propped up against legacy networks. Sonet becomes unnecessary in an all-IP world, because packet protocols like IP thrive even when lower layers are unreliable. ATM loses when Internet telephony and audio-on-demand thrive, because more bandwidth and a few IP tweaks promise to make real-time and streaming media scream.

To St. Arnaud, the whole idea of convergence is a backward-looking attempt to preserve existing assets. He proposes a divergent, third residential network for Internet traffic only, installed alongside telephone and cable feeds. Like the CA*Net 3 backbone, it’ll have only two layers: IP and WDM. Information over light. It’ll be a stupid network — cheap and simple, under-engineered, over-provisioned and controlled at the edge by users.

Gigabits for microcents Installation (right-of-way, trenching and conduit) represents the most cost. In a 100-km. metropolitan network, a conservative installation estimate is $4.3 million. Routers and equipment to light the fiber might cost another $1.8 million. Using today’s 128-wavelength equipment, a single 48-fiber cable would serve 6,144 homes. Each home would have its own WDM wavelength that could be lit at 2.5 gigabits per second (OC-48). This computes to $1,000 per home.

The alternative, new hybrid fiber-coax (HFC) to support cable modem service, delivers hundreds of times less, and costs half again more. Even retrofitting existing cable to carry two-way data could cost $600 per home. DSL, somewhat cheaper, delivers even less than HFC.

CANARIE’s optics would meet residential equipment at an Ethernet interface. The step from 2.5 gigabits (OC-48) to 1 gigabit Ethernet might seen wasteful. But St. Arnaud points out that the next Ethernet evolution — 10 gigabit Ethernet — just happens to match OC-192. Local and wide area nets would merge in yet another fundamental simplification.

Why Canada can In Canada, a lot of municipal fiber already exists, thanks to favorable regulatory policy. But in the U.S., bean-counters of communications behemoths shy from huge installation costs. They look at today’s applications and figure that current networks can be kludged to handle them.

Make way for high-definition Internet video on demand, or whatever truly broadband application Canadian users dare discover. CA*Net 3 could make Canada the center of the next Internet economic boom. Meanwhile, U.S. telcos manage mawkish mergers, dither with DSL and forget fiber to the home. Look north, young entrepreneur.

Copyright 1999 Advanstar Communications.