Blog

Taking turns can slow you down

As discussed in my last article, when deploying Wi-Fi networks there are a number of considerations around wireless connectivity that we need to be aware of when comparing them to the more familiar world of wired networking.

In this article we’re going to look at another of Wi-Fi’s “guilty secrets” that few people seem to be aware of. Are you ready for it..?

“Wi-Fi is a shared medium”

What does this mean? All devices connected to each wireless access point (think of laptops, iPads, smartphones etc.) have to take turns to transmit and receive data. They have to share access to the access point. This means that as more devices that join an access point and have data to send, the longer each device may have to wait before it can send its data.

Wi-Fi is a “contended” medium: everyone using it has to take their turn to “talk” to the wireless access point. When things start to get busy on your Wi-Fi network, this can be a significant challenge.

Why is this an issue compared to wired (think Ethernet) networks? Well, each device connected to an Ethernet switch has its own dedicated chunk of network access bandwidth. If an Ethernet station is connected to a 1Gbps port, then it has a full 1Gbps available and does not share it with anyone else. Each device connected to an Ethernet switch port gets its own 1Gbps of access bandwidth to use.

In a Wi-Fi network, if we have a single device connected to an AP, it has access to the full bandwidth that the AP makes available. If we add a second device, the bandwidth available is now shared between both devices (50% bandwidth each) as they take turns to send data. If we add a third device to the AP, each device now has access to 33% of the bandwidth as three devices now take turns to send data. As more devices are added to an AP, the opportunities for each one to send data will decrease, reducing the bandwidth available to each one.

Effect of Multiple Wireless Clients Contending For Wi-Fi Access (Simplified)

The contention to access the wireless network described above applies only if multiple devices need to send data at similar times. However, if wireless forms a significant part of the access layer of your network, this will very likely become an issue very quickly as the number of users (who may have several Wi-Fi devices each) increases.

How do we mitigate the limitations of Wi-Fi’s contended access method? We must optimize our wireless design to achieve the maximum throughput available. This is achieved by designing each Wi-Fi network for capacity as well as wireless coverage. Careful design & planning of the wireless RF environment, together with taking account of client & application bandwidth requirements is the only way to ensure optimized performance.

No matter how many gazillions of dollars, pounds, euros or <insert your currency here> you spend on your Wi-Fi network, it is subject to the same immutable laws of physics as everyone else’s. This applies to home networks, mid-sized and even large Enterprise Wi-Fi networks.

Wi-Fi has a number of inherent limitations, compared to wired networks, that are (currently) insurmountable and limit the data throughput that can be achieved. Wi-Fi networks generally achieve a fraction of the throughout that you would expect from a modern switched, wired network. Unless you understand these limitations and design with them in mind, your wireless users are likely to have a poor experience.

The first limitation of Wi-Fi networks that we’ll look at is the fact that it uses a “half duplex” medium. This means that if your iPad, iPhone, Android tablet, laptop (in fact any client device using a Wi-Fi connection) is exchanging data with a wireless access point (AP), data flows in only one direction at any time.

Data frames can flow from the AP to the client device, or vice versa, but never in both directions at the same time. This limitation is due to the fact that the wireless AP and associated clients all use the same channel to communicate, meaning that they can only be transmitting or receiving.

Half Duplex vs Full Duplex

This contrasts sharply with wired networks, where full duplex is the general default mode of operation. This means that a device connected to an Ethernet switch port may have data frames flowing simultaneously in both directions. This effectively provides twice the available throughput for an Ethernet connected device, compared to a Wi-Fi connected device with a similar connection speed.

How does this translate in to the real world? If we consider a wired device connected to a 1Gbps Ethernet switch port, it can realize a data throughput of close to 1Gbps connection speed, if required. If we have a Wi-Fi client connected to an AP using 802.11ac at a physical connection speed of 1.3Gbps, we’re likely to achieve an actual data throughput of around half the connection speed (approx 700Mbps), due to the half duplex nature of its connection.

Unfortunately, Wi-Fi clients that can connect at 1.3Gbps are in the minority, with many 802.11ac mobile clients only able to support lower speeds. Also, there are still many legacy, lower-speed, devices that may be in use on your Wi-Fi network which provide even worse performance.

Even if you are happy with your 700Mbps throughput, there are more factors to consider which mean that even this level of performance is a pipe-dream in the real world. We’ll look at more of these in future posts.

How do we mitigate the limitations of Wi-Fi throughput? We can’t. What we must do is optimize our wireless design to achieve the maximum throughput available. Careful design and planning of the wireless RF environment is the only way to ensure optimized performance.

There is often confusion around which channels may be used when deploying a Wi-Fi network using the 2.4GHz band.

Depending on where you are in the world, the band generally has either 11 channels (numbered 1 to 11) or 13 channels (numbered 1–13) available to use.

Unfortunately, it’s not as easy as just using any channel that you might feel inclined to use. In short, the convention is generally to use channels 1, 6 and 11.

Why not the intervening channels (you may ask)?

The 11 (or 13) channels available on the 2.4GHz band are each 5Mhz apart. Wi-Fi communications requires a channel width of 22MHz to operate (the equivalent of 5 channels). We cannot use channels that are adjacent to each other, as we would get an effect called “adjacent channel interference”. This is a very bad thing in Wi-Fi networking, leading to very poor network performance.

You may have heard that is best for your network to use its own dedicated channel, which is true (when feasible). But if you find that neighboring Wi-Fi networks are already using channels 1,6 & 11, do not be tempted to try to use intervening channels for your network (e.g. 2, 7, 12). Co-existence on the same channels as nearby networks is better than trying to use other channels.

The generally accepted convention is to use channels 1, 6 and 11. I strongly advise that you stick to the same convention.

(Side note: You really should be leveraging the 5GHz band as far as possible for Wi-Fi networks — 2.4GHz is generally not a good choice in many instances in terms of Wi-Fi network performance)