Multi-Gigabit AP Backhaul - Do you need it?

I was recently asked by a Wi-Fi engineer about the potential need for multi-gigabit backhaul needs from an AP with the pending release of 802.11ac Wave 2. This seems to be a point of confusion for many in the wireless industry. Here's what I told them:

Industry claims of throughput capabilities exceed 1 Gbps are correct from a theoretical standpoint. However, real world client mixes on almost every WLAN will mean that backhaul never approaches even close to 1 Gbps of throughput.

First, when you combine clients of varying capabilities there is no chance of exceeding 1 Gbps backhaul. The only time you will need more than 1 Gbps of backhaul is in POC bakeoffs between vendors, lab tests, and very low-density locations where you have only a few users on an AP radio but they are using top of the line high-end wireless laptops and applications that can push large amounts of data (I'm thinking CAD users here for instance who collaborate and push files of several GBs across the network and want it done fast). This is somewhat counter-intuitive because most people would think off-hand that high-density areas is where you'll need the greater backhaul. But in high-density areas you run into the client mix which consumes more airtime at slower speeds as well as more contention related overhead. 

Second, 160 MHz channels still won't be possible there just aren't enough channels in an enterprise environment for re-use. So the top-end speeds won't be available.

Third, clients will almost never have good enough signal to achieve the highest 256-QAM modulation rate. So the top-end modulation rates won't be achievable.

Fourth, even IF (and I stress IF) you assume that 160 MHz channels can be used, the noise floor is higher as channel widths get wider and the required signal strength that a client device requires in order to achieve the highest modulation rates is higher. Realistically to achieve the highest 256-QAM modulation rate with a 160 MHz wide channel you're looking at needing an SNR in the range of 40 dB! If the noise floor is at -90 dBm that would be -50 dBm... insane! Sorry, this just isn't going to happen. More realistically, you would see high-end clients data rates top out around 1.5 Gbps instead of the 3.466 Gbps that will be advertised by manufacturers (4SS, 160 MHz). And I've already seen tons of media coverage around Wave-2 quoting the 6.9 Gbps theoretical maximum speed when 8SS are supported - sorry to burst everyone's bubble but Wave-2 APs will be 4SS and we won't get to 8SS for a long time (if ever).

Fifth, many people get confused about MU-MIMO... so let's set the story straight. No - this is NOT turning a Wi-Fi AP from a hub into a switch. Bad analogy. What it does do is allows the AP to use otherwise unused spatial streams if it's transmitting to clients with lower capabilities. So, if a 4SS AP (which is likely what we'll see with Wave-2) is transmitting to two laptops that are 2SS each, then it can utilize its extra spatial streams to serve multiple clients at once. So while it is increasing aggregate WLAN capacity and throughput over time, it is not affecting the top-end speed or peak backhaul out of an AP. You will still be able to pump the most instantaneous throughput when a single high-end client (3SS or maybe soon 4SS) is connected and there is no medium contention.

So, I would only advise customers to think about >1 Gbps backhaul with Wave 2 in very specific locations that have a low-density of high-end devices with very demanding bandwidth needs. Plus that area has to be fairly well isolated RF wise so that you can actually use 160 MHz channels. For these very rare situations, 10 Gbps links aren't the solution - PoE isn't supported yet, it requires CAT6 cabling, APs likely won't support 10Gb ports since they cost a lot more, and fiber plus AC-electrical outlets are a pain and costly. I also don't like the idea of running two cable drops to an AP either, unless you have other reasons to do so (security segmentation, failover, etc.). So realistically you're looking at using the new Multi-Gig technology such as NBase-T from Cisco that runs at 2.5 / 5 Gbps over CAT5e and supports PoE+ and UPoE (60W).

Long story short: don't upgrade cabling if it's CAT5e or better and take a hard look at the locations you think you need NBase-T to verify you actually need it. Chances are you don't.

BTW [shameless self-promotion] - you can use my Capacity Planner tool to model client mixes and the resulting backhaul traffic load on APs very easily. I'm not sure if you have seen it but it's available at 

For instance, even if you look at a scenario with a high-end 4SS laptop and 160 MHz channel width, here is what you could expect to get when the client achieves a realistic data rate of 1.56 Gbps (SNR ~26 dB) - around 866 Mbps throughput. Remember that the 2.4 GHz radio doesn't support 11ac since its a 5 GHz-only protocol, so you get 11n on 2.4 GHz with a 20 MHz channel. Plus I'm being considerate here assuming a 160 MHz channel width, a 4SS client that doesn't yet exist, and zero medium contention.

And if for some reason you still aren't convinced, go read what my great friend and Wi-Fi rockstar Marcus Burton wrote on this same subject.