Ethernet, the data networking protocol often hailed as the most ubiquitous enabling technology in the communications sector, celebrated its 40th birthday a few weeks ago. It was invented and developed at PARC – Xerox' legendary Palo Alto Research Centre – by a team of engineers led by Bob Metcalfe. Metcalfe had been tasked with devising a simple, yet flexible, networking scheme that would link several hundred pcs and some printers at a distance of a mile and operate at several hundred kbit/s – and with a minimum of cabling, to get rid of the 'rat's nests' of wiring common at the time.
The 'team' included a bunch of young, highly motivated researchers who Metcalfe called the 'unsung heroes' of Ethernet. It included, most importantly in view of his huge input into the project, David Boggs, a Stanford graduate who is credited on the Ethernet patents, together with Chuck Thacker, responsible for much of the diagnostics, and Tat Lam, who devised the first Ethernet transceivers. "The list goes on, but must include Ron Crane, who subsequently designed the first working 10Mbit/s coax based Ethernet and who played a most significant part in standardisation efforts through the IEEE802.3 group." Metcalfe recruited Crane as the hardware guru at 3Com, the company he founded in 1979 to take Ethernet to market. Metcalfe's memo outlined his vision of Ethernet, drawing on the AlohaNet concept, an innovative radio based network of computers he studied as part of his Harvard doctoral thesis. "Incredibly, we had packets flying by 11 November," said Metcalfe. In this first working prototype, a 2.94Mbit/s CSMA/CD (carrier sense multiple access with collision detection) network connected more than 100 workstations on a 1km cable. The more pragmatic and practical Boggs still maintains the November date is the real 'birthday' – when Ethernet was actually functioning. "It has been reinvented many times since over the past 40 years, operating now at speeds approaching 100Mbit/s and with work already pushing that to 400Mbit/s," noted Metcalfe. And Ethernet is used over all mainstream media, including twisted pair, fibre and – of course – wireless. Metcalfe added he would be the first to admit that not much remains of the original concept, save for the name and, importantly, the basic 'type' frame. While collision detection may have left the scene – replaced by network intelligence at level 3 – the main message is that the packet format has stood the test of time, as has the addressing scheme – even with switched Ethernet. "The technology remains robust and, most importantly, is backward compatible," stressed Metcalfe. "It was important then and is huge now as we see data rates pushing into the terabit range and bandwidth demand just exploding." It is what has allowed the technology to be adopted in a vast range of networks, from industrial control to core communications networks, where it is replacing Sonet. It underpins Wi-Fi and cellular backhaul, has spawned Carrier Ethernet and enables end to end networking. And it was revealed last month that Ethernet is heading to the cloud. Metcalfe referred to a '$100billion market' with some 1.2bn Ethernet ports shipped last year, 800million of them wireless, according to market research group IDC. Not surprisingly, there have been several skirmishes along the way. Having left PARC and founded 3Com, Metcalfe manoeuvred hard to get patent holder Xerox to jointly promote Ethernet with what became known as the 'DIX consortium', including Digital Equipment and Intel, and to make it an open standard. There were other LAN specifications eyeing the same space. In the blue corner, was IBM (Big Blue) with Token Ring – which appeared to have an advantage because it was based on twisted pairs – and Token Bus, a proposal from General Motors that was soon seen off. It was a long and sometimes acrimonious battle played out in the market and in standards bodies. "IBM quite often played a dirty game", recalled Metcalfe. "But we won out, the market won and open standards won." Some of the reasons? Ethernet was cheaper and easier to deploy. Metcalfe noted: "Because, from the very beginning, it was designed as a layer 1 and 2 protocol, it was comfortable in that space and understood its role in the communications hierarchy." It helped that, in 1982, several companies, including important players in Europe, agreed to use a particular Ethernet specification in their devices, even though it had yet to be approved by standardisation setting groups such as the IEEE. Metcalfe, meanwhile, is now in his fifth career, but declined to tell New Electronics which gave – or gives – him the most satisfaction. "I try not to compare them, but in each what I love most is the learning curve. And I am definitely in one now as a Professor." Metcalfe joined the University of Texas Austin's faculty of engineering as Professor of Innovation in late 2010, with the aim of helping to train engineers to start and grow their own businesses. No doubt he stresses to students what he has preached throughout his career – 'nothing happens until something gets sold'. He noted that he made no money from the actual invention of Ethernet, rather through the arduous work of selling products based on the technology while at 3Com. And the real money came when HP acquired 3Com – half jokingly, Metcalfe suggests it was the other way round. He left the company in 1990 when he lost a boardroom battle to Eric Benhamou to lead the company he founded. He took a year out, accepting a fellowship at the University of Cambridge, then returned to the US to become an outspoken columnist (with his regular 'From the Ether') and pundit, conference organiser and, later, publisher and executive at IDG's Info World. Ten years later, he joined the rush into venture capital, becoming a General Partner at Polaris Venture Capital, investing and helping the development of numerous start ups in the semiconductor and networking sectors. "I also broadened my interest in and passion for the opportunities in the green energy, healthcare and education sectors," he added. Clearly, Metcalfe has grabbed every opportunity and headline during his multiple careers. Who would bet against him in his quest at Austin? And more importantly, who would bet against Ethernet? Creating the pathways towards a 400Gbit Ethernet One important message from the recent Ethernet Innovation summit was that the real strength of the networking protocol was not what has been achieved over 40 years – but what it could, should become. There was no hint of any mid life crisis – instead, the focus was on the crucial role Ethernet can play in meeting the insatiable appetite for capacity and bandwidth and higher speeds. Yet the calls to start considering Terabit rates were balanced by economic and technical realities and the fact that 40 and 100Gbit/s products were only just starting to be rolled out and that the high volume market for 10Gbit has yet to be reached. John D'Ambrosia, chairman of the Ethernet Alliance and the chief Ethernet evangelist with Dell, is the man just tasked by the IEEE to define a version of the technology capable of 400Gbit/s. He suggested 'we have reached a point of inflexion'. Speaking to New Electronics, D'Ambrosia said the group will be facing multiple challenges, but also huge opportunities to shape the next generation. "We will have to keep in mind that solutions need to be driven by economics; in today's climate, companies have to justify deploying networks, so the engineers' deliberations will need to focus on solutions that allow the cost per bit to fall." So what will a 400G Ethernet look like? "We only had our first meeting last month so it is too early to be specific. We are now awaiting the proposals." And the timescales? The former chair of the 40 and 100Gbit/s standardisation effort noted that this took some four years for ratification. Because the group will clearly be at the forefront of both silicon and photonics technologies and will have to cater for numerous port configurations and link lengths, D'Ambrosia said: "I suspect it will take about the same time to reach consensus." D'Ambrosia said there are a host of options in all the different categories that have to be specified. "From an electrical signalling perspective, even though advanced modulation was used in Base-T signalling for years, it is likely that other modulation schemes, perhaps NRZ types, will be needed. We are also likely to need novel types of forward error correction (FEC). All this will play a bigger role in deploying Ethernet across copper and thus extending the life of copper. But we may need to devise an FEC that can be embedded into one of the layers." There will also need to be much thought about providing full support for Optical Transport Networks, while energy efficiency will need to be improved. This latter point could also become a major marketing tool. A big debate is likely to follow on the best architectures for 400Gbit, ones which could adapt to evolution in signalling technologies. For example, one able to adapt from 16 lanes at 25Gbit/s to 8 lanes at 50Gbits/s to 4 lanes at 100Gbits/s to a single layer at 400Gbit/s would seem to a reasonable expectation. "But a 16 x 25Gbit/s solution feels sort of ugly," said D'Ambrosia. Yet a huge investment has already been committed to interfaces that divide signals into 16 lanes, for instance the 300pin transponder and numerous optical components. For cost reasons, the ecosystem is likely to be based on 100Gbit multiples – so would a 16 x 25Gbit solution be a short term solution? And how many short term solutions can there be? "These are the complex, multiple decisions that are keeping me awake at night," quipped D'Ambrosia. He is clear that, from an optical signalling perspective, there will be a multimode option, but what to expect in terms of reach is less obvious. As well as playing a vital role in Ethernet standardisation efforts, D'Ambrosia, as noted, is an evangelist for the technology. As such, he is pushing at every opportunity for another IEEE802.3 project that would define Gigabit Ethernet delivery over something less than the standard four pairs used today – perhaps one or two pair. One obvious application would be Ethernet connectivity in vehicles, where the port count could become really significant. And then, he mused, why not a small form factor Ethernet for the consumer sector, for instance in tvs, hand held devices? If any of that sounds fanciful, D'Ambrosia revealed that, within the IEEE802 executive committee, there is already talk of a whole range of consumer oriented applications, such as lighting and amusement parks.