Data Center

The Ethernet We Know Today is 35 Years Young

by Steve Day on ‎09-30-2015 08:00 AM (3,515 Views)

Although it was originally conceived in 1973 at Xerox Parc, it was the 1980 specification jointly issued by Digital, Intel and Xerox on September 30th that really got the ball rolling on the Ethernet we know today. Yes, it was “only” 10Mbps. Yes, it used “vampire” taps. But for its time it was progress. It was good enough and cheap enough to get people interested in using it rather than the alternatives of the day.

 

But it would have faded out had it not shown the ability to evolve several aspects when necessary: transmission medium, link connectivity, and speed. Critically, that evolution happened just in time for the early majority of the target markets, and there are workarounds for early adopters, e.g. link aggregation groups, or accept the risk of using pre-standard specifications.

 

800px-Diffusion_of_ideas.jpg

The diffusion of innovations according to Rogers (1962). Source: Wikipedia

 

The transmission medium changed from “thick” co-axial cable, to “thin” co-axial cable, to twisted pair and fibre. These changes made Ethernet easier to install, easier to change, and easier to accommodate the range of higher speeds to come. Today it is commonplace for domestic consumers to build simple networks using twisted pair Ethernet cables.

 

The link connectivity evolved from a length of cable with vampire taps or T-piece connectors, to hubs, and then switches. These too helped with easier installation and change, but also allowed the aggregate bandwidth across all end-points to dramatically increase. This was because Ethernet evolved from being a shared medium to being a dedicated medium for each end-point. This is still an unavoidable challenge for many other communications media in use today e.g. Wi-Fi, mobile/ cellular.

 

10base2_t-piece.png

 

A 10Base2 T-Connector. Source: Wikipedia

 

Last, but by no means least, is speed. 10M, 100M are legacy. 1G, 10G, 40G, 100G are mainstream. 2.5G, 5G, 25G, 50G are emerging. 400G is in specification development. Originally that speed applied to a shared medium, but by the time 1G was reached the speed applied to point-to-point links. Ultimately this is what network operators want: the careful mix of link speeds which lets them provide the aggregate network capacity to meet business goals at the right price over the required geographic area.

 

Note that there has been little change in Ethernet frame formats. This has provided vital stability for operating system driver code, and switching and MAC ASIC designs. Otherwise these factors would have been major barriers to migration from one generation to the next. 

 

None of this would have been possible without some other things being in place. Firstly, there is an industry-wide consensus  that creating and adopting specifications with broad support is mutually beneficial. For example, the Ethernet Alliance provides a technology roadmap. Secondly, the semiconductor industry is the epitome of continual innovation and of the benefit of economies of scale which can be passed on to network operators i.e. affordability. Finally, the physicists who have pushed our understanding of signal transmission over copper and optical fibre i.e. speed over distance. 

 

To sum up, Ethernet is 35 years young and as vital now as it was when it was 25. It’s easy to use, affordable, adaptable and thus ubiquitous. If only the connections for consumer electronics had had such a record, but that’s another story.