12 Jan 1.2 The Story of Technology

Nature’s physical limits are not easy to overcome. Certain laws of physics are literally sacrosanct as engineers have been butting their heads against them for decades without much success. During my technical career, I experienced the thrill of discovering engineering breakthroughs as well as the futility of overcoming Nature’s physical limits.

The mid nineties were heady days in the history of the Internet. The Internet community was confined mainly to technical users then, but we were laying the foundations for the popular version of the Internet that emerged at the turn of the century. In those early days, the large transnational telecommunications corporations were huge supporters of a tiered data communications system called Asynchronous Transfer Mode (ATM), where all data transmission requests were to be sent to a central server, which would then determine precisely how the data gets routed from source to destination over the Internet. But the small Silicon Valley companies were promoting a peer-to-peer, distributed networking technology called Ethernet, where any computer server could plug in and participate as an equal in the routing of data, in keeping with the nascent principle of net neutrality of the Internet. As you can imagine, if the ATM protocol had won the data communications war then, the Internet would have been very different and John Oliver wouldn’t be railing against the Federal Communications Commission (FCC) in 2014 as it tries to eradicate net neutrality[19].

Net neutrality would probably never have existed at all!

Indeed, by late 1995, the ATM Forum had standardized on a 155.52 Mb/s (million bits per second) data rate for its fastest links on copper wires, while the Ethernet standards committee had standardized on a slower 100 Mb/s data rate.  Yet the freewheeling Ethernet protocol was being rapidly deployed in the marketplace due to its inherent simplicity and due to the robustness of its slower 10Mb/s copper links. But there was a hitch. The newer 100Mb/s Ethernet links were experiencing intermittent failures, causing server connections to break down and resulting in customer complaints. The chairman of the Ethernet committee at that time was worried enough to ask me, as a systems specialist, to take a look at the protocol and suggest ways to improve it.

The next committee meeting was to be held on Jan. 8, 1996 and I had a flight to catch in the morning of Sunday, Jan. 7 to make my presentation to the committee. Standards committee work is meant to be done pro bono and I normally gave it lower priority than my day-to-day work that paid our bills. Thus, as usual, I spent the evening before the flight at my office in New Jersey studying the problem and preparing for the presentation. To my incredible delight, I discovered that while the 100Mb/s protocol had some niggling issues in the specifications, the physical wiring itself was capable of easily supporting a data rate that was TEN times faster, i.e., 1000Mb/s or 1 Gigabit per second! I was so sure that the committee would be thrilled with that discovery!

Late that night, after finishing up my presentation, I turned off the lights in the office and pushed on the front door to go out. It wouldn’t budge! It was completely blocked by wind drifts from the snow that had been falling through the evening and I was stuck in the office. Overnight, a major nor’easter dumped 14 inches of snow in our area and up to 48 inches in other places and the whole state of New Jersey was shut down. This was the first and only night that I slept on the couch at the office. Early next morning, my wife, Jaine, called the maintenance crew and had them clear a narrow pathway to our office door first so that I could get to Newark airport on time. I still recall the eerie feeling as I rushed between two huge, ten-foot tall walls of snow on either side of me in front of the office, to catch my flight to make my presentation the next day.

At the end of the presentation, everyone in the room laughed! The chairman of the committee said,

“We’re having trouble getting 100 Megabit to work and you are telling us that we can crank it up 10 times faster? Dream on! I’ll believe it when I see it.”

Nevertheless, he was kind enough to let the committee flesh out these ideas. It turned out that my calculation on that eerie snow-bound night was real, not a fantasy. A concrete proposal that I developed a few months later became a standards protocol.

The resulting 1000BASE-T standard was the first widely deployed communications standards protocol that relied on high-speed Digital Signal Processing (DSP) technology for Ethernet as opposed to the 100BASE-T standard that could be implemented with standard analog techniques[20]. 1000BASE-T was adopted as a standard in 1999 and during the heydays of the Internet mania, annual shipments were over 150 million units by 2003, within 4 years. The digital 1000BASE-T technology developed a reputation for being more robust on the same cabling than any analog 100BASE-T device and it still forms the backbone of the Internet infrastructure. The wired network connection on your laptop is most likely a 1000BASE-T Gigabit Ethernet link.

Starting in 2003, I began attending the 10 Gigabit Ethernet Task Force meetings, where a more sophisticated DSP technology, 10GBASE-T, was being discussed to propel data ten times faster than 1000BASE-T over similar copper wiring. The same chairman who had told me to “Dream on!” seven years earlier had now become a wide-eyed techno-optimist and he was clearly expecting 10GBASE-T to be a piece of cake based on his 1000BASE-T experience. However, this time around, we were reaching physical capacity limits on the wiring and I knew that it was not going to be easy at all. Nevertheless, the Ethernet standards committee approved the 10Gigabit Ethernet on Copper standard, 10GBASE-T, in 2006 almost unanimously. I stopped working on it, partly because I felt that it would not meet the high expectations of the user community, but mainly because I became interested in environmental issues instead. Five years after that standard was approved, the worldwide shipment of 10GBASE-T totaled a mere 182,000 units[21]. According to the Linley group, electromagnetic interference issues and large power consumption were cited as the main reasons for the poor uptake of 10GBASE-T devices.

As of now, no one is talking about 100Gigabit Ethernet over copper cabling. The exponential Ethernet speed growth over copper cabling is well and truly over. Nature sets physical limits and we have no option but to abide by them.

Therefore, it is time to stop expecting technological miracles to prolong the exponential growth phase of the human enterprise.

---

[19] John Oliver’s net neutrality segment can be found here: https://www.youtube.com/watch?v=fpbOEoRrHyU

[20] 1000BASE-T has now become a commodity item that can be purchased as a core for integration in larger devices:

[21] 10GBASE-T shipment numbers for 2011 can be found here. . The poor uptake of 10GBASE-T on existing cabling has prompted the development of 2.5GBASE-T and 5GBASE-T variations under the NGBASE-T alliance.