The coming of Wi-Fi 6 does not mean it's time to ditch your cabled LAN. Here's why
There's a time and place for wireless tech, but it still can't compete
Feature IEEE 802.11ax-2021 (more commonly known as IEEE 802.11ax or, more familiarly "Wi-Fi 6") was approved on 9 February 2021, with a top speed of 1.2Gbit/sec per single stream (think "stream" as synonymous with "channel"). As seems to happen each time a new Wi-Fi technology comes out, people are yet again asking whether this is the one that will finally tip us over the edge and entice us away from cables and onto wireless.
I'm going to stick my neck out and ask a slightly different question, and explore whether we're ever going to move our worlds to dispose with wires.
First of all, let's not get bogged down with the potential of moving the server room and the data centre into a wireless world: that's simply not going to happen, ever. In the average server room or data centre that uses physical servers, each box will have at least a pair of Gigabit Ethernet connections linked redundantly into the switch infrastructure. Gigabit Ethernet is full-duplex (strictly speaking the original spec included half-duplex, but it never really went anywhere) and so as long as the devices at each end can keep up, you get pretty much a gigabit constantly (and I say "pretty much" only because you have to allow for the bits of the Ethernet frames that aren't the data payload, such as the headers).
In a virtualized world where you have your servers running virtually on, say, Hyper-V or EXSi hypervisors, 10 Gigabit Ethernet is the common choice – again paired up for resilience.
And the thing is, it's actually relatively cheap to do cabled networking in the server room or data centre. Whatever technologies you run, there is always going to be some Layer 2/3 hardware at the core and (if you have one) at the access layer to fling the traffic the right way. In server environments the cabling requirement is in-cabinet at best, and at worse a few bits of copper or fibre string between cabinets that are adjacent to each other. Cabling your office can be expensive (hold that thought for later) but cabling the data centre isn't.
So, with Wi-Fi just about getting somewhere near 10Gbit/sec (using multiple streams) at a time when bits of wire into servers can run at 40 times that speed, it's hard to envisage how wireless will ever come close to catching up. We'll revisit shortly what makes a long-term cabled server room a certainty.
What about out there on the office floor?
As hinted at, the main motivation for wanting to go wireless is cost. It's some years since I rolled out the LAN of a new premises from scratch, but I recall the structured cabling bill for the last one I did – a big but not vast single-floor office – coming in significantly north of £100k. And complexity shouldn't be underestimated either: one must not forget to factor in the time and effort taken to patch floor ports to PCs, and comms room structured cabling outlets into switches, and the need over time to ensure that any changes are recorded diligently.
It should also be pointed out that there are some instances where Wi-Fi is a bit of a no-brainer. This correspondent used to be a judge of the UK Networking Industry Awards, which ran for some years from the mid-1990s. I remember distinctly an award entry in the "Project of the Year" category, in which an IT service provider (now defunct) had been faced with the challenge of implementing a corporate network in its listed-building headquarters. The solution to the problem was predominantly Wi-Fi, and the company was delighted with the results. You hear such stories from time to time, and they do sow a seed of "maybe" in one's mind.
- Network managers think you're coming back to the office. Why else did they go on a Wi-Fi 6 buying spree?
- Ethernet standards wonks eye up speeds beyond 400Gb/s
- Latest phones are great at thwarting Wi-Fi tracking. Other devices, not so much – study
- Wi-Fi 6 isn't signed off yet, but boffins are already teasing us with specs for venerable wireless tech's next gen
Back to cabled networks, though: if they're so eye-wateringly expensive, do they bring any real advantages? Well, the benefit Ethernet brings by the truckload is guaranteed bandwidth: when your laptop fires up its Gigabit Ethernet port it has that full-duplex path to and from the switch and can squirt at will. Next in line is the fact that Wi-Fi doesn't play nicely when there are walls, doors, steel building supports or frankly anything solid between the user's PC and the wireless Access Point (AP). No amount of cunning design will enable Wi-Fi traffic to walk unimpeded through solid things. And finally, Wi-Fi is – and always has been – a half-duplex broadcast technology, where the endpoint has to watch and wait for a clear spot before transmitting rather than (as is the case with full-duplex networks) just sending.
Not a switch
There is a school of thought that questions this latter point, observing that 802.11ax is pretty much a switch. But although one can see the point being made, it's not quite the case. Tom Hollingsworth, otherwise known as The Networking Nerd, pointed out almost two years ago in his fairly firmly titled blog article "802.11ax Is NOT A Wireless Switch" that no matter how much tinkering has happened, Wi-Fi is still not proper full-duplex. As Hollingsworth puts it: "Importantly the transmission medium still hasn't magically become full-duplex. Stations may get around this with some kind of trickery, but they still need to wait for the all-clear to send data. Remember that all stations and APs still hear all the transmissions. It's still a broadcast medium at the most basic."
Does this mean that Wi-Fi can never be a switch, and hence never run at a guaranteed, full-duplex speed? Well, in its basic form one would think that full duplex should be possible: one can imagine an AP and a user endpoint having a pair of dedicated radio channels – one sending and one receiving at each end – for their exclusive one-to-one use. But there are problems here.
First is that we have to differentiate between stuff that moves and stuff that doesn't. The main benefit of Wi-Fi is that we can wander around the building with our phones and tablets, move from our desk to a meeting room and remain on the network, yet what we don't see is that the endpoint and the APs seamlessly negotiate with each other to find the most suitable AP that will give the best experience. The dedicated channel approach just doesn't work in that context, so you'd end up with a hybrid world where some of your kit – the devices that move – would have to go with the traditional model of negotiating a connection with whichever radio it's able to talk to.
Next is the attenuation issue: Wi-Fi only performs to the max of its speed rating when there's a relatively short distance and (as we've noted already) nothing to soak up the signal between source and destination. Cables – once you've run them over, under or through the walls, of course – simply don't have that issue.
Third is interference: we've all seen videos on the web of microwave ovens interfering with Wi-Fi networks in the 2.4GHz band, but at any frequency there is the potential for electromagnetic (EM) interference on wireless networks, including the company across the courtyard cranking its AP radio signal strengths up to 11 so the radio waves spill in through your windows. LANs made of fibre-optic cables are immune to the levels of EM interference in the average office, but even copper-structured cabling behaves very well when there is interference about. I spent my most geekfesty day ever some years ago at ITT Cannon's EM testing lab in the south of England, in fact, and I was surprised just how resilient even unshielded twisted pair (UTP) cabling is to EM interference. Tests on FTP cable (Foil Twisted Pair – like UTP but with metallic foil wrapped around) showed it to be utterly unbothered by whatever EM rubbish we bombarded it with. (I also learned, incidentally, that companies that ban mobile phones from their comms rooms are missing a trick if they don't also ban their security guards' way more powerful walkie-talkies.)
Don't cross the streams
So will Wi-Fi ever triumph? The clincher is this: Wi-Fi works within limited frequency ranges, and there will always be a finite number of streams (channels) available. Once you hit that limit, devices have to start sharing – so the designers' only option is really to make the streams narrower and narrower and produce radios with really strict tolerances that don't let the channels trample on each other. This constant narrowing of streams simply can't happen ad infinitum. In this sense, the limitation in a cabled network is effectively infinite – it's limited only by the number of bits of electric or glass string you choose to install.
Wi-Fi will continue to develop, then. And rightly so: Wi-Fi is the perfect tool for people who move around buildings, who undock their laptops from their (Ethernet-connected) desks and gather in meeting rooms. But it is just not possible to imagine a world where Wi-Fi will, as a mainstream technology, reach a state where it's as fast and bullet-proof as cabled networks. And yes, cabling is expensive to do it but because you have to re-cable so seldom, if you amortise the cost sensibly, it fits comfortably within the average budget.
We will, therefore, continue our approach of cabling our desks and using Wi-Fi where convenience demands it. The cabled LAN will take the strain when we dock our laptops, and leave the (faster and faster) wireless bandwidth for the people whose devices can't, or don't want to, plug in. ®