More About Performance
In the previous blog post about SSIDs, I talked about how management traffic can consume the airtime and slow your WiFi network performance. This post addresses similar performance impact created by poor channel use. I’ll attempt to explain why the number of channels, and their size, can have a big impact on the WiFi.
Remember this from my WiFi Airtime post..?
..when the signal drops too low, devices may not receive all of the data being sent, or some of the data be become corrupt. In either case, the data will need to be resent to the receiving device (increasing airtime utilization)..Teaching WiFi Airtime – Part 2
There are several reasons why a WiFi transmission (or frame) must be resent, and we’ll discuss another one here in a moment. What triggers these retries is the absence of an acknowledgement (ACK) frame from the receiving device. You see, a lot can go wrong with WiFi, so the powers-that-be built into the protocol this acknowledging behavior where the receiver will acknowledge the successful receipt of a transmitted frame, by sending its own ACK frame. The scary thing is that these retries are resent over-and-over (and usually slower-and-slower) when the expected ACK isn’t received. What do you think that does to the available airtime?
A Bit about WiFi Channels
Most already know that WiFi operates in the 2.4 and 5 GHz frequencies, and there are a finite number of non-overlapping channels for use in those spaces. 2.4 GHz has just 3 non-overlapping channels (1, 6, and 11), and 5 GHz has 20-ish channels (ruling out the four channels that can be impacted by Doppler weather radar). Actual available channels may vary by vendor.
With the introduction of 802.11n, we were given the ability to combine two channels together to double the connected speed, or data rate. This is commonly referred to as channel bonding, or increasing the channel size or width. One channel equates to a width of 20 MHz, and two bonded channels equates to a width of 40 MHz. Later, 802.11ac gave us the ability to bond four (80 MHz), and even eight (160 MHz) channels, quadrupling and octupling the speed. Okay, that sounds pretty great. The faster our data is transmitted, the less airtime is used. No brainer, right? Well…
Why Channel Size Matters
Math tells us that as we double the size of the channels, we also halve the number of channels available for use. In 5 GHz for instance, you get:
- # of 20 MHz channels = 20
- # of 40 MHz channels = 10
- # of 80 MHz channels = 5
- # of 160 MHz channels = 2
Now, let’s say you are in change of a facility that has 25 APs deployed, and you are using 80 MHz wide channels (because that’s fast). This means that each channel must be reused five times. When you reuse channels, you have a much greater likelihood for co-channel interference. Co-channel interference can slow your network down as devices must wait longer for their transmit opportunity, and it may also result in corrupted frames. Corrupted frames can’t be interpreted by the receiver, so the receiver won’t send back an ACK, which we now know results in the transmitter re-sending the frame over-and-over until it gets an ACK, or gives up (resulting in packet loss).
So, using wide channels decreases available channels, which can increase co-channel interference, and reduce network performance. Strike one!
Big Channels, Big Noise
Another impact of using wider channels is that there is more noise (ambient background radio energy). Again, as you double the channel size, you also double the amount of background noise, which then increases the noise floor, and has a direct impact on the all important signal-to-noise ratio (SNR).
[You see, when an AP and its connected devices communicate, they must distinguish signal from noise, and if the noise floor rises, it’s much harder for them to do that.]
Specifically, the noise increases by 3 dB, so the SNR at the client device drops by 3 dB. If the channel width is 80 MHz, then the noise increases to 6 dB, and so on. WiFi Nigel wrote about his experience testing this theory, and you can read about it here.
Another way to look at this is from the perceptive of cell size, or the distance from the AP where you will get a particular signal level. When designing a WiFi deployment, you typically design per the requirements of the client devices that will be in use. If a client device requires a minimum SNR of 25, you can measure how far you can be from the AP upon reaching that level. Assume that using 20 MHz wide channels, you can have an SNR of 25 at a distance of 40 feet. However, if you are using 40 MHz wide channels, and considering its increased noise, you may hit an SNR of 25 at 30 feet. Your AP cell size just shrank by 25%. Ouch!
More noise, and therefore less SNR – that’s strike two!
Speed vs. Capacity
It’s true that bonding channels gives us faster connected speeds. But, that doesn’t mean that our WiFi network will then be better capable of servicing more client devices effectively. I know, you’re probably thinking, “What’s wrong with you Todd? The faster the speed, the quicker a client device will finish transmitting, and the sooner the next client device can start transmitting. So yes, faster speeds will allow you to service more clients.” And yes, you are right, but think of it this way. Let’s say you have one AP using a fast 80 MHz channel, covering a classroom of 50 client devices. All of the devices need to transmit data to the AP, and they are transmitting very fast, one at a time. Now, let’s say that instead of one AP at 80 MHz, you have four APs on 20 MHz wide channels. Now, your WiFi network can actually service four clients. Even if these clients are transmitting slower than those connected to the AP using the 80 MHz wide channel, the four APs are still able to service more client devices more quickly than the one faster AP.
I’m going to borrow an analogy from Devin Akin. He related this to the line to use the bathroom during a break in a class he was teaching. The nearest bathroom only had one toilet, and there were many people waiting in line to use it. Now, if people used the toilet in the absolute shortest amount of time possible, there would still be a long line, but it would move along a little quicker. However, if the bathroom had four toilets to use, imagine how much quicker the line would move.
I recommend that everyone reading this post also listen to Devin relate this experience himself, where he also talks about this same subject to much greater detail. Listen to the great “Clear To Send” podcast here, where Devin was the guest.
Faster speeds, yet less network capacity. That’s strike three channel bonding, you are out! In most cases, using wider than 20 MHz channels just doesn’t make much sense.