What is Ethernet? The standards, explained
We take a deep-dive into the backbone of modern networking
While far from the most exciting topic, Ethernet technology is crucial to the way devices such as computers are connected to the internet. The technology, and the standards underpinning it, are essential to keeping the world connected.
Although Ethernet is complex, especially for those not versed with the ways of the networking world, understanding how ethernet connections work is crucial to holding down a role in IT. Taking this information on board, moreover, offers a lot that can be used to solve problems relating to faulty networks.
This level of complexity exists throughout the IT world, of course, with technology and the standards that underpin them absolutely crucial. Unfortunately, they’re always a nightmare to understand and quite difficult to get your head around on occasion. These are also largely academic and difficult to relate against real-world examples.
USB is a perfect example of this. When the USB 3.2 standard was released, a full rebrand ensued that saw the slower USB 3.0 and 3.1 connections redefined into USB 3.2 Gen 1 and USB 3.2 Gen 2. USB 3.2, meanwhile, will now be known as USB 3.2 2x2.
The key theory behind Ethernet is the TCP/IP protocol technology first proposed by Robert Metcalf in his PhD thesis in the 70s. This technology was then standardised in a patent filed by Xerox, who employed him, and cited David Broggs, Chuck Thacker and Butler Lampson as coinventors. They were true pioneers, and this moment represented the first time that such technology had been created. To make matters tricky, however, there was no foundation to build this framework on top of because nothing like this had ever existed. This is why the documentation reinforcing Ethernet technology is so chunky.
The history of Ethernet development
When Ethernet was first devised, the technology comprised coaxial cables of various grades which one was expected to pierce with a long spike. They're referred to more commonly as drop leads or patch leads, which were large and stiff 15-pin connectors. This was quickly revised, with a breakthrough coming in the form of a more slender coaxial, with twist-link barrel connectors for joints and wall sockets.
At such an early stage there was no need to tweak any elements of the software or recompile programmes to transition from one wiring standard to the other, which some argue come from a wider-cast standard of work around the seven-layer OSI transport model. Although this could very much be the case, it doesn't take into account a basic Ethernet fundamental from one of the earliest standards. This is that Ethernet actually doesn't demand any specified layout of a packet. Different frame types with varying functions can be left open to implementation.
The implementation's openness allowed one to expand the standard to a high degree from the beginning. It's a particular concern for businesses, with this becoming a key reason your IT expenditure could potentially balloon. With plenty of buildings floor-wired to service users with Ethernet via unshielded twisted pair (UTP), it can be a real benefit but also something of a drawback. This is because it's easy to install a wiring plan that turns out to be disabled by standards as opposed to enabled by them.
This is because the success of UTP as a business networking format has been so immense that early and quite rigorous adherence to standards has been left in the dust. Gone are the days when every single piece of wire in a whole installation would be accompanied by a time-consuming and exhaustive signal quality report, or indeed when you would see qualified technicians laying cable in carefully routed trays.
In some cases, you'll find straightforward electricians putting up Cat5E cables with cable retaining nails, neatly smashing into the solid conductors inside the UTP enclosure - naturally these looked fine while machines were connecting at 10MB/sec but utterly refuse to function with gigabit ethernet over UTP requirements. The new standard requires all eight conductors within the cable sheath to function perfectly to spec, which rapidly exposes hasty or ignorant installation practices.
Ethernet standards explained
|Status||Officially phased out in 2003 - heavily superseded|
|Effective range||500m depending on number of drilled holes|
|Uptake||Early years only but quite widespread|
|Presentation||Thin coax, 50 ohm, barrel twist connector|
|Status||Phased out in 2011 - switchless and therefore all about collisions|
|Effective range||Varies according to installation|
|Uptake||Very widespread through 1980s. Pivotal tech for Novell LAN cards|
|Lifecycle||Still live in a few places but ripe for replacement|
|Presentation||Unshielded twisted pair, 8 conductors. Hub-centred|
|Status||Almost universal - used by faster standards too|
|Effective range||100m from switch or repeater|
|Uptake||Key network enabler for most of the planet|
|Presentation||Optical fibre (2 strands)|
|Status||Sometimes found in early campus wiring, telecoms et cetera - 2km range remains useful, and effort of laying cables may be impossible to repeat|
|Uptake||Small, as fibre is expensive and difficult to maintain|
|Presentation||Unshielded twisted pair, 8 conductors, hub-centred|
|Status||Almost universal - desktop data transfer fits perfectly|
|Effective range||100m over copper|
|Uptake||Cat5E cabling is most common (though other variations may be found)|
|Presentation||Unshielded twisted pair, 8 conductors, switch-centred|
|Status||Server rooms, switch links, desktops - not all cable installs support Gbit speed well|
|Effective range||100m over copper|
|Uptake||Still not as widespread as it could be|
|Presentation||Cat6A twisted-pair, switch-centered|
|Status||Mostly server rooms - different wiring pays back with higher speeds|
|Effective range||25m to 400m according to installation|
|Uptake||Slow, as many find it hard to realise the speed benefits|
25, 40 & 100BASE-T
|Presentation||Cat8 cabling at a minimum|
|Status||Server data centres, storage arrays - all are subsets of the 100Gbit project|
|Effective range||30m (single lane) up to 30km (4-lane fibre)|
|Uptake||Low, as deployment is complex and specialised|
|Lifecycle||Likely to be long but only in specialised deployments|
The future of ethernet
It's easy to assume that the ramp of both speed and sustained capability is likely to keep on increasing: 10Gb Ethernet was implemented to function over common Cat5E, though those who draw attention to this as proof of universal compatibility, clearly haven't tried to actually make that work in practice.
The move to 10GbE has been accompanied by a bit of a first: Dropping of some previously universal and reliable standard features. No half-duplex, no Collision Detection: If this seems concerning then very possibly you are not best placed to be pushing out 10GbE, or some application you are using has been a trifle cavalier with their implementation (although, to be fair, this is mostly a VOIP telephony problem).
Those who follow the speed race assiduously will be sneering. The leading edge of both standards and hardware development is currently hovering at the 100Gb mark: Why on earth would anyone with a job to do actually refuse to go faster?
Because faster ethernet is delivered with an ever-increasing series of restrictions and ever-rising costs. Before gigabit ethernet was developed for UTP copper cables, it was readily available so long as you were ready for optical fibre cabling in a business environment. Fibre's main advantage in this context has been that it is largely immune to bodgers: Fibres have to be handled carefully, routed sensitively, and terminated cleanly and with surgical, submicroscopic precision.
You can run an ethernet fibre for 70km a satisfying upgrade from the usual UTP limit of 100 metres, to be sure but that involves some considerable expenditure and work way outside the usual network engineer's remit, of establishing things like wayleaves and the geological profile of the terrain.
Most of the faster platforms for connected devices by Ethernet especially the popular 40Gb standard used in many datacentres and supercomputer builds make not just extensive use of fibres; they make use of multiple fibres, bundled into a cabling format that is so far from the DIY-friendly simplicity of UTP that it puts an end to casual, thrown-together networks forever, at those higher speeds.
It really is important not to fret about lost performance opportunities in business ethernet deployments: For every user who thinks that high throughput is the same thing as low latency, there are justifications for keeping desktop traffic speeds down. It's green, for example: Modern switches adopt another standard which turns down the power when traffic is light.
It's also not really necessary if you have shifted to a Cloud or Thin Client computing model. A sensible Cloud or Thin Client session is very bitty (that is, it ships data in very small chunks), and quite low volume: Imagine a gigabit ethernet session carrying typing keystrokes, each one within a surrounding padded-out 512 byte minimum packet size. That padded-out behaviour doesn't apply at 100mb, making the slower session, faster.
Only by looking at the standards and understanding their intentions, can you get a rational business justification for going faster, or indeed, sometimes, slower.
Why use Ethernet?
Given that wireless connectivity is a staple of most devices today, you may wonder why anyone needs to know about Ethernet beyond its role as the backbone of a business network. The reality is that Ethernet still occasionally finds a use, largely due to the limitations of wireless technology.
A business Wi-Fi network is not always the easiest thing to deploy, especially in larger offices with lots of corridors, walls, and multiple floors that will block or hinder connectivity. This is especially noticeable when a business has a hotdesking policy, as the quality of the connection can vary considerably as you move around a building. To keep some semblance of a consistent connection, it’s common for office spaces to still offer a wired Ethernet link to the network.
It’s not just the quality of the connection, either. Wired connections generally deliver much faster, and far more consistent, speeds than the typical wireless connection, even if you’re sitting next to the router.
In theory, recent innovations in wireless technology should make wired connections obsolete. Wi-Fi 6, for example, should be a match for any speed you would get through a wire. Yet, wireless connections are still at the mercy of the environment they’re deployed in, and until thick walls and floors no longer hinder connections, Ethernet will always have a place.
Managing security risk and compliance in a challenging landscape
How key technology partners grow with your organisationDownload now
Evaluate your order-to-cash process
15 recommended metrics to benchmark your O2C operationsDownload now
AI 360: Hold, fold, or double down?
How AI can benefit your businessDownload now
Getting started with Azure Red Hat OpenShift
A developer’s guide to improving application building and deployment capabilitiesDownload now