Living legends: John Cioffi, father of DSL

Cioffi says he became fascinated by the DSL project, and 30 years later he's still at it

John Cioffi

John Cioffi was there when Bell Labs decided in the 1970s to turn the single remaining analog part of the AT&T network - the twisted-pair copper wires that run to each customer's home - into a digital connection; 30 years later he's still at it.

Watch a timeline of networking in the past 25 years.

Credited with being the father of digital subscriber line (DSL), Cioffi says he became fascinated by the project, and that hasn't flagged over a career that includes a long stretch as a professor at Stanford University as well as heading up two DSL startups.

The initial goal was to support high-bandwidth video services on last-mile phone lines, but that proved elusive for more than a decade. Initially, digitizing the local loop resulted in ISDN BRI lines that support two voice channels or a single 128Kbps data channel, not enough for good quality video.

Finding a modulation scheme that would support higher bandwidth was what Cioffi worked on for awhile, but he moved away for a few years to IBM before signing on as a professor at Stanford in 1986 where he resumed his DSL research. "I got serious about it, and we achieved data rates for good video," he says -- 1.5Mbps or more coming from the network to the customer. In 1991 Cioffi felt he had made enough progress to take a leave of absence from Stanford to start Amati, one of the early companies to make DSL modems and that was bought by Texas Instruments in 1997.

His chief accomplishment was developing and championing discrete multi-tone modulation (DMT), also known as bit swapping, which is the scheme used in asynchronous DSL (ADSL), the most popular commercial flavor of the technology. DMT breaks down the frequencies used on the twisted wire pair into groups called carriers, and further divides the carriers into smaller segments called subcarriers. Data itself is divided into separate streams called channels that are sent, one each, down the subcarriers.

When DSL modems based on DMT train on each other at opposite ends of a copper pair, they test the capacity of each subcarrier, which can vary greatly from pair to pair depending on the length of the line, its condition, signal interference from nearby copper pairs, AM radio signals and electromagnetic emissions from appliances. During their initial engagement, the modems determine the best loading of each subcarrier to result in the best throughput for the line at that time.

"Every phone line is different, and DMT is a more adaptive approach," Cioffi says. "It can find the sweet spot for each line, and they're different with different phone lines - your line, your neighbor's and the guy down the street."

Read Network World's 25th anniversary package

Demand for inexpensive broadband services blossomed with the rise of Internet use in the mid-1990s, and DSL services became common. Cioffi and Amati rode that wave, and when he returned to academia after selling Amati, Cioffi maintained his interest in DSL. This time he researched management software that would maximize DSL capacity across entire carrier networks, not just over individual lines. His efforts resulted in the development of dynamic spectrum management (DSM), which boosted DSL throughput to 100Mbps over some 1 km lengths of twisted pair and even to 1Gbps under ideal conditions.

The goal of DSM is to milk the highest overall aggregate bit rate out of DSL lines that lie close enough to have an impact on each other's performance. So DSM might be used to figure out that lowering the transmission power of a single customer's DSL modem could significantly boost the bit rate on 10 other DSL lines in the same binder by reducing the interference on those 10 lines. DSM works in several ways, starting with monitoring network parameters including actual bit rates vs. maximum possible bit rates, errors and forward error corrections. Analysis of those factors results in tuning individual pairs of modems to achieve overall optimum throughput for the network.

In 2002 Cioffi says he had completed enough research to start up a company called ASSIA to write DSM software for service providers. He claims ASSIA manages 80% of the 35 million DSL lines in the U.S. Worldwide, 50 million of the 350 million DSL lines in service are under the management of DSM, he says.

DSL so far has managed to stave off competition from fiber-to-the-home, which promised broadband access speeds right away and virtually unlimited bandwidth over time without replacing the fiber itself. Verizon's FiOS program - replacing copper lines with fiber - is the prime example in the U.S.

Continuing demand for broadband plus the expense of fiber programs and improvements to DSL mean that DSL deployments continue. "You can't afford fiber everywhere," Cioffi says. "Copper is lower cost and it can provide the bandwidth immediately."

It appeared that fiber networks or cable TV networks would outstrip DSL for awhile, but worldwide the fiber efforts have waned - China is the only country pursuing an all fiber last mile, with the government calling for an end to DSL. Despite the 2009 edict to halt DSL deployments there, they have continued to grow from 80 million that year to 100 million in 2010. "Even a communist edict couldn't stop DSL," Cioffi says, because it's still the fastest and least expensive way to meet bandwidth needs.

Cioffi says he is optimistic that DSL will continue to grow, surprisingly fueled in part by wireless technology, Wi-Fi-enabled phones in particular. Consumers connect these phones to Wi-Fi networks at home to cut costs, and their local Wi-Fi is by and large connected to the Internet via DSL, he says. About half of U.S. homes with Wi-Fi connect to the internet via DSL, he says.

That encourages Cioffi to stay with DSL for a few more years at least. When he started out, a colleague at Bell Labs told him it would take 30 to 40 years to digitize the local loop. Since then 30 years have gone by and Cioffi sees another 40 years before the goal is reached.

Did you ever have an A-HA! moment while developing your tech?

I think it was more accumulation of various discoveries but some important ones were the recognition around 1987 that DSL would require an "adaptive multi-carrier" solution (the "DMT with bit-swapping" I mentioned), the invention of vectoring in 2000 (recognizing that crosstalk noise between lines could be removed), and the creation of the concept of Dynamic Spectrum Management roughly around 2000.

Did you think your technology would have the impact that it has?

Not really as while I was an early user of email/Internet (as early as 1978-1979 for email), its impact was not really foreseen by anyone and of course DSL has become by far the main deliverer of the Internet.

What technologies do you think will be important in the future?

The sharing of Wi-Fi and small-cell wireless connections by users at the ends of the network to billions of portable and stationary application devices (smartphones, tablets, ultimately dozens of communicating appliances) at the ends of the DSL connections that will carry the load for the next 40 years (or maybe longer, again, after this first 40 years my friend projected).

Learn more about this topic

Network World's 25th anniversary package of stories

Slideshow: Networking during the past 25 years

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Now read: Getting grounded in IoT