Imagine a lecture hall that can hold four hundred students. Then imagine that the lecture hall has been tested for 5 Gbps aggregate throughput. Even at two and a half devices per student, that should be more than enough. Back-of-the-napkin math says that's at least 5 Mbps per device.
Now imagine that the students are in the lecture hall and the Wi-Fi stinks. "It's slow," they say. A peek at the controller shows under 100 Mbps for all the APs in that lecture hall, combined. How is that possible? How can 5 Gbps turn into 100 Mbps so quickly?
The answer is interference. Interference is the common cause of the all-too-common problem of tested wireless networks failing to perform to expectations.
Knowing that interference can cause Wi-Fi problems is the first step. That first step, however, is the easy part. Things get more difficult when things need to get more specific. Hard questions need to be answered: What is causing the interference? Can the interference be avoided? Will solving the present interference problem create new problems somewhere else? This paper aims to help in answering those questions.
- TABLE OF CONTENTS
- STEP ONE: Identify non-Wi-Fi interferers
- STEP TWO: Locate non-Wi-Fi interferers
- STEP THREE: Identify Wi-Fi interferers
STEP ONE: Identify non-Wi-Fi interferers
It is best to start an analysis with non-Wi-Fi interferers because non-Wi-Fi interferers can make good Wi-Fi impossible. If, for example, a hospital has a DECT phone system (DECT being a separate, non-Wi-Fi technology), it can kill Wi-Fi across the entire 2.4 GHz frequency band. (Modern versions of DECT use different frequencies than 2.4 GHz, but the problem could exist around older DECT systems.) That is because DECT phones don't share. When a DECT phone needs to use wireless, it uses wireless. And many other wireless technologies—from Bluetooth to zigbee devices—work the same way: no sharing.
Wi-Fi interferers are almost always less harmful than non-Wi-Fi interferers because they share. Wi-Fi devices use 802.11 contention, which causes devices to listen and check the channel before transmitting. That checking and listening means that Wi-Fi devices tend to share channels tolerably with each other. In almost all cases, non-Wi-Fi devices do not share as well because they do not use contention.
What, then, can be done about non-Wi-Fi interferers? The best bet is to identify them, locate them, try to determine their impact and then adjust accordingly.
First: identify the interference source. The key is to emulate the experience of the end user as Wi-Fi devices generally have internal radios. If a spectrum analyzer has a similar antenna as a Wi-Fi device, then interference seen in the spectrum analyzer is more likely to be seen by the end user's Wi-Fi device as well. Spectrum analyzers that operate using external antennas may detect interference that does not affect end-users. Investigating interference with no impact wastes valuable troubleshooting time.
STEP TWO: Locate non-Wi-Fi interferers
Once a non-Wi-Fi interferer has been identified, it must be located. Once the device has been located, it can be dealt with according to the location's policies. Ideally an interfering device can be disabled, but that is not always possible.
STEP THREE: Identify Wi-Fi interferers
Once non-Wi-Fi interference sources have been handled, then nearby Wi-Fi devices should be analyzed.
Now, knowing what to look for. Wasted channel time is the greatest killer of Wi-Fi performance because channel time is the one resource that is limited. The number of packets on a wireless channel can increase: if there are fewer packet errors, then there can be more packets. The amount of data (bytes) on a channel can increase: if data rates improve, then more data can be accessed by each device. But one second is one second. If one second —or part of one second—is lost it cannot be recovered.
There are several ways that time can be lost. Collisions cause the wireless channel to lose time because data that suffers a collision has to be sent again. That means that the initial transmission of data was a waste of time. Low speeds waste time as well. Data rates are measure by taking data and dividing it by time. If the data rate is lower, then that just means that it will take more time to send the same amount of data. Non-data traffic can also be a waste of time, if it is unnecessary.
First: Collisions. A collision is a failed Wi-Fi data transmission. The 802.11 standard (the standard that Wi-Fi is built on) specifies that if data is sent and an acknowledgment is not received (thus indicating that a collision happened), then the device or AP that sent the data must mark the retransmitted data as a Retry. This means that the percentage of Retry data is equal to the percentage of data transmissions that suffer collisions.
In order to put collision statistics to good use, one has to know what is considered a high Retry percentage. A good place to start is at 8% for ordinary Wi-Fi and 20% for difficult Wi-Fi (high density of users, lots of mobility or significant amounts of non-Wi-Fi interference). Once Retry percentages climb above those numbers, it is usually a good idea to set aside some time and investigate why so many retransmissions are happening.
The second big time-waster is low data rates (often called “speeds”). Data rates can be seen in the same general areas that Retry percentages can be seen. The only difference is that in order to see which data rates are being used, the “Speed” tree needs to be expanded after clicking on an AP or station device in the Infrastructure screen.
Determining whether low data rates are a fixable problem can take time and careful analysis. Devices and APs—especially 802.11n/ac devices and APs—routinely use data rates well below their stated maximum rates, even if interference is not significant. In other words, an office worker with an 802.11ac smartphone (maximum rate on a 40 MHz-wide channel: 200 Mbps) connected to an 802.11ac AP might routinely use data rates below 150 Mbps, even if the RF environment is great. 802.11ac (and, to a lesser extent, 802.11n) includes a lot of technologies that are rarely available during typical enterprise usage, even in areas where interference is minimal. For that reason, analyzing data rates to determine interference problems is a task that usually requires some experience.
Thirdly (and lastly), non-data traffic can be the cause of a Wi-Fi channel losing time. There are lots of different types of non-data traffic, but most of them are mandatory for 802.11 operation, and thus cannot be eliminated. APs and stations have to exchange a number of types of non-data traffic to be able to stay connected, detect collisions and all sorts of other necessities for a wireless network to function.
There are, however, two types of non-data traffic that can sometimes be reduced: Beacons and Probes. Beacons are used by APs to let stations know that a Wi-Fi network is available. The problem is that each Wi-Fi network needs its own set of Beacons. If there are two SSIDs (one guest and one internal), then Beacons will typically take up between 2% and 5% of the available time on a channel. But if the number of SSIDs expands to eight (possibly by having unique SSIDs for different vendors or different groups of internal users), then Beacons will typically take up between 8% and 20% of channel time. That's a big difference. And the problem gets exacerbated if more than one AP is covering the same channel. If a given enterprise tablet can see three APs on channel 11 and all three of those APs are using eight SSIDs, then that makes for 24 sets of Beacons on the channel. That’s 24% to 60% of your channel time that is being used for instead of for Data.
Probes are the other type of non-data frame that can waste channel time. It’s important to identify whether they are causing a problem. If there is a Probing problem, try making sure that the station in question has a stable Wi-Fi connection. Modern Wi-Fi devices (smartphones, tablets, laptops, etc.) do very little probing if their Wi-Fi connection has stable access to the Internet.
Handling Wi-Fi interferers has some similarity to handling non-Wi-Fi interferers, but there are also big differences. It is always best to start by identifying and locating the interferer. After a Wi-Fi interferer is identified and located, the problem can often be minimized or eliminated by adjusting the wireless LAN infrastructure. Disabling AP radios, adding new APs in different locations, manually configuring AP channel numbers and setting AP transmit power to levels similar to those of client devices are all activities that can improve an infrastructure of APs and controllers. In contrast, non-Wi-Fi interferers often need to be disabled or avoided in order to get the Wi-Fi working.
Following these steps creates an excellent chance of preventing Wi-Fi interference from becoming a lasting problem. While these steps may take some getting used to, it sure beats blind exercises in trial-and-error, and can make all the difference in successfully identifying and resolving interference problems.