Thursday, 15 May 2014

802.11ac & 5GHz: The Emperors New Clothes? - Part 3

After previously looking at the challenges we may face with 802.11ac due to the restrictions of the 5GHz band (see part 1 and part 2 of this series), in this final installment, I suggest how we may mitigate the challenges we may face, together with a possible (but no doubt controversial) solution.

The Solution?

I'm not aware of any "magic bullet" to solve all of the challenges outlined in this series of articles (sorry). More spectrum will certainly help things. But, time-scales for the allocation and adoption of new spectrum are not clear at this point in time, though many agencies around the world are considering ways to free up more spectrum to improve WiFi capacity.

Though spectrum may become available in relatively short time-frames, there will still, no doubt, be a significant time-lag before this permeates through in to the world of vendor and consumer WiFi products (remember the issues raised around UNII-2e support?) 

The best way to mitigate the issues discussed in this series of articles is through:
  • a thorough understanding of the clients that will be using a wireless LAN - for instance, do they support UNII-2e/DFS channels, do they support 802.11ac, how many spatial streams do they support, what channel widths do they support?
  • careful planning of the channels to be used throughout a facility, taking account of channels used, channel widths and DFS considerations
  • a good wireless design/survey taking account of channel planning and co-channel interference/contention - this is more important than ever!
As I said, there's no magic bullet - just careful consideration of the points raised in this series of articles, which ends up being a series of carefully planned compromises due to the challenges that 5GHz presents.

Many of the points raised in this article have been an issue since the introduction of 802.11n, with it's 40MHz channels and the increased emphasis by vendors and manufacturers on 5GHz. However, there is now even more focus on the 5GHz band (specifically) with the introduction of 802.11ac. I'm still surprised how many people who I speak to who are still unaware of the 5GHz band for WiFi usage, so there are plenty of folks who are still new to all of this information.

Many folks have likely adopted default settings for their WiFi equipment without really giving too much consideration to channel planning. Many will have accepted a default of perhaps 8 x 20MHz channels operating on channels 36 - 64, which has worked just fine for them. With the advent of 802.11ac, this approach no longer suffices - more thought and planning is required if the true benefits of 802.11ac are to be realized.

To wrap up this discussion, 802.11ac is not simply about a wholesale migration to the 5GHz band to deliver ubiquitous Gigabit WiFi connectivity for your organisation. It requires careful planning and consideration (despite what the marketing boys may tell you).

Depending on your client devices and attitude to (DFS) risk, you will be somewhere on a spectrum of marginal improvements over your existing WiFi solution, to amazingly fast WiFi. To quote a well know policeman: "’ve got to ask yourself one question: 'Do I feel lucky?' Well, do ya, punk?"

An Alternative?

There is an alternative approach that I would not be surprised to see some manufacturers start to more seriously consider: single channel architecture (yep...there, I said it).

In a single channel architecture (SCA) solution, all APs and clients operate on the same channel. Note that this is used by a wireless systems specifically designed to operate in an SCA mode, you can't just flip all of your existing APs on to the same channel.

Some manufacturers already provide an SCA solution, though they are currently regarded in some quarters as being something of a 'niche', 'non-standard' solution. However, they are certified by the WiFi alliance, so this  is something of an unfair label to apply to them.

There is no doubt (in my mind anyhow) that a SCA can never perform to the same level as a traditional, well-designedwell-deployed 'multichannel' architecture (MCA). If you start to consider the amount of capacity available to each non-overlapping cell in a MCA solution, the maths speak for themselves - the capacity is proportional to the number of non-contending cells available in a network. However, there are three key assumptions underlying the MCA vs SCA argument:
  1. There are a reasonable number of channels available to allow an MCA solution to be deployed
  2. All AP cells in an MCA are sufficiently isolated from other AP cells on the same channel that they do not suffer co-channel interference/contention (hence cutting the potential capacity of two cells by a significant factor)
  3. The MCA solution has been well designed and deployed (i.e. cell sizing, channel planning and co-channel contention considerations are all well implemented)
Given the limitations of the 5GHz band once we start to consider the wider channels that 802.11ac brings, and the fact that very few WiFi networks are particularly well designed in the real world (in my experience), then I believe that the SCA approach has significant merit.

Although it will not be able to hit the potential theoretical throughput that an MCA solution could provide, it can provide a ubiquitous solution that uses an 80MHz channel on non-DFS channels - a feat that cannot be achieved by MCA systems.

I believe it has more chance of delivering the promise of 802.11ac, as it allows the use of the wider channels and the accompanying benefits. Many MCA solutions will have to compromise by reducing channel widths and negate a significant advantage that 802.11ac delivers.  Though multi-user MIMO should certainly deliver increased throughput capabilities in environments with many single-stream devices, the underlying channel width issues will remain. 

WiFi gurus out there will no doubt be able to plan and tweak their WiFi networks to the nth degree and get a super-high throughput network operating on MCA. If they can find organisations who will pay for that level of expertise and commissioning (and are able to maintain their network), that's great. They will no doubt achieve excellent results.

For the vast majority of everyone else deploying WiFi, with customers who want a solution in a working in a timely and affordable manner, then it might be time to at least consider an alternative (and for vendors to start providing that option). SCA certainly isn't the solution for every environment, but is worthy of consideration in many scenarios.

I know that the proposition I'm putting forwards here won't be popular with many WiFi professionals, but I think that SCA is at least worthy of consideration as an option by more WiFi vendors. 

If you've been in the networking industry for a while, you might remember two competing LAN technologies: token ring and ethernet.

Token ring was faster, providing 16mbps LAN speeds, compared to the paltry 10mbps of Ethernet. It used an efficient, well-ordered token passing mechanism, compared to the rather inefficient and erratic random back-off and collision detection methods of Ethernet.

In many ways, token ring was a  technically superior, better-performing solution than ethernet.

It's flaw: it was complex, expensive to maintain and difficult to deploy in comparison to the ease and simplicity of rolling out UTP hubs to create an ethernet LAN. Despite its advantages, people simply got fed up with the complexity and expense of deploying and maintaining token ring networks, resulting in Ethernet winning the LAN technology battle. 

Sound familiar..?

(Thanks to Andrew (@revolutionwifi) for allowing me to use the image featured in this article. It is taken from his original article: 802.11ac Channel Planning