Adding APs to an ESS
If a client is associated with an AP, it can maintain the association as long as it stays within range of the AP. Consider the cell shown in Figure 7-5. As long as the client stays within points A and B, three conditions are met:
- The client is able to receive the AP’s signal at an acceptable level.
- The AP is able to receive the client’s signal.
One of the acceptable modulations can be successfully used between the client and the AP.
Figure 7-5 A Mobile Client Moves Within an AP Cell.
As soon as the client goes outside the cell range at point C, one or more of the conditions fails and the client loses the association. In the figure, the AP’s signal has fallen below an acceptable threshold.
Other APs can be added so that the client can move within a larger area; however, the APs must be carefully deployed to allow the client to roam from AP to AP. Roaming is the process of moving an association from one AP to the next, so that the wireless connection is maintained as the client moves.
In Figure 7-6, a new AP has been added alongside AP-1, each using the same channel. It might seem intuitive to build a larger coverage area by using a single channel. Usually this turns out to be a bad idea because the client may experience an excessive amount of frame collisions in the area between the two cells.
Figure 7-6 Pitfalls of Reusing Channels in Adjacent Aps.
Remember that the signal from an AP does not actually stop at the edge of the cell; rather, it continues to propagate as it eventually dies off. This is shown by the signal strength graph of each AP. The client is able to form an association with AP-1 at point A. Even at that location, some portion of AP-2’s signal can be received, albeit at a lower level. Because AP-2 is using the same channel as AP-1, the two APs (and any clients within range) can essentially interfere with each other through co-channel interference.
Ideally, when the client in Figure 7-6 moves to location B, it should begin to anticipate the need to roam or transfer its association from AP-1 to AP-2. Notice that AP-1 and AP-2 are spaced appropriately for roaming, where their cells have some overlap. The two APs are out of range of each other, so they are not aware of each other’s transmissions on the same channel. Each AP will coordinate the use of the channel with devices that are inside its own cell, but not with the other AP and devices in the other cell. As a result, the client around location B will probably experience so many collisions that it may never be able to roam cleanly.
The Roaming Process
What enables a client to roam in the first place? First, adjacent APs should be configured to use different non-overlapping channels. For example, an AP using channel 1 must not be adjacent to other APs also using channel 1. Instead, a neighboring AP should use channel 6 or higher to avoid any frequency overlap with channel 1. This ensures that clients will be able to receive signals from a nearby AP without interference from other APs. As you learned in Chapter 2, “RF Standards,” the 5-GHz band is much more flexible in this regard because it has many more non-overlapping channels available.
The roaming process is driven entirely by the wireless client driver—not by the AP. Wireless clients decide that it is time to roam based on a variety of conditions. The 802.11 standard does not address this at all, so roaming algorithms are vendor specific. In addition, the roaming algorithms are usually “secret recipes,” so the exact thresholds and conditions are hidden from view.
Some of the ingredients in the roaming algorithm are the received signal strength indicator (RSSI), signal-to-noise ratio (SNR), a count of missed AP beacons, errors due to collisions or interference, and so on. These are usually logical choices because they indicate an inferior connection.
Because different clients use different thresholds, some will try to roam earlier than others at a given location within a cell. Some clients will tend to “latch on” to an existing association until the AP can hardly be heard, whereas others will attempt to roam whenever a better AP is discovered.
Figure 7-7 depicts a clean roam between two APs that have been correctly configured with non-overlapping channels 1 and 6. The two AP signal strengths are also shown as a graph corresponding to the client’s location. At location A, the client has a clear signal from AP-1, so it maintains an association with that AP.
Figure 7-7 A Client Roaming Correctly Between Two APs.
As the client moves toward location B, it decides that AP-1’s signal is no longer optimal. Somewhere along the way, the client begins to gather more information about any neighboring AP cells. The client can passively scan by tuning its radio to another channel and listening for beacons transmitted from other APs. During the time that the radio is tuned away from the associated channel, the client might lose packets that have been sent to it. A client might use active scanning instead, where it sends probe requests to seek out a better AP where it can move its association. The client does not know what channel is used on the next AP it encounters, so it must send the probes over every possible channel. Again, the client must take time to tune its radio away from the current AP’s channel so it can scan other channels and send probes.
You might think of this as someone watching television. As the current program gets boring or nears its end, the viewer begins to “channel surf” and scans other channels for a better program. One thing to keep in mind: While the viewer is scanning channels, he cannot keep watching the original program. Some of that program will be missed. This is also true of wireless clients. While a radio is scanning other channels, packets arriving on the original channel will be dropped because they cannot be received. Therefore, there is a trade-off between staying available on a single channel and attempting to roam to other APs.
After the client is satisfied with all of the beacons or probe responses it receives, it evaluates them to see which AP offers the most potential for a new association. Returning to Figure 7-7, when the client nears location B, it receives a probe response from AP-2 on channel 6. At location C, the client sends a reassociation frame to AP-2 and moves its association to that BSS.
How much should cells overlap each other to promote good roaming? Cisco recommends 15 percent to 20 percent overlap for most applications. The idea is to give a client device some continued coverage even after the RSSI of its associated AP falls below a threshold and a roam might be triggered. The client can probe and reassociate with the next AP before it completely loses contact with the previous AP. Seamless roaming is especially important for time critical applications like voice traffic.
WLAN Channel Layout
The previous section laid the foundation for roaming by describing movement between two AP cells. Most scenarios require more than two APs to cover the appropriate area within a building. Therefore, you need to consider the layout and configuration of more and more APs to scale the design to fit your wireless environment.
For example, to cover the entire area of a warehouse or one floor of a building, APs must be placed at regular intervals throughout that space. A site survey is a vital step toward deciding on AP placement, as actual live measurements are taken with an AP staged at various points in the actual space. This method also takes any factors like free space loss and absorption into account, as the signal strength is measured within the actual environment where clients are located.
To minimize channel overlap and interference, APs cells should be designed so that adjacent APs use different channels. For simplicity and a convenient design constraint, the examples in this section use the three non-overlapping 2.4-GHz channels. The cells could be laid out in a regular, alternating pattern, as shown in Figure 7-8.
Figure 7-8 Holes in an Alternating Channel Pattern.
However, notice what is happening in the center where the cells meet; there is a small hole in RF coverage. If a client roams through that hole, his wireless signal could drop completely. In addition, if the cells were brought closer together to close this hole, the two cells using channel 1 would overlap and begin interfering with each other.
Instead, you should lay the cells out in a “honeycomb” fashion, as shown in Figure 7-9. This pattern is seamless, leaving no holes in coverage. In addition, notice how the two cells using channel 1 are well separated, providing isolation from interference. As far as ordering channels in the pattern, there are several different variations using combinations of the three channels, but the result is basically the same.
Figure 7-9 A Better Alternating Channel Pattern.
Notice that as the client shown in the channel 1 cell moves around, it will roam into adjacent cells on different channels. For roaming to work properly, a client must be able to move from one channel into a completely different channel.
Alternating channels to avoid overlap is commonly called channel reuse. The basic pattern shown in Figure 7-9 can be repeated to expand over a larger area, as shown in Figure 7-10. Naturally, this ideal layout uses perfect circles that are positioned regularly across the building. In practice, cells can take on different shapes and the AP locations may end up being irregularly spaced.
Figure 7-10 Channel Reuse Over a Large Area.
So far, only the channel layout of a two-dimensional area has been discussed. For example, Figure 7-10 might represent only one floor of a building. What happens when you need to design a wireless LAN for multiple floors in the same building?
Recall that an RF signal propagating from an antenna actually takes on a three-dimensional shape. With an omnidirectional antenna, the pattern is somewhat like a donut shape with the antenna at the center. The signal extends outward, giving the cell a circular shape along the floor. The signal also extends upward and downward to a lesser extent—affecting AP cells on adjacent floors as well.
Consider the building with three floors shown in Figure 7-11. The same two-dimensional channel layout from Figure 7-10 is being used on the first floor. The floors in the figure are shown greatly separated, so that you can see the channel patterns and numbers. In reality, the cells on adjacent floors would touch or overlap, just as adjacent cells on the same floor do.
Figure 7-11 Channel Layout in Three Dimensions.
The pattern of alternating channels exists within the plane of a floor and between floors. Channel 1 on the first floor should not overlap with channel 1 directly above it on the second floor or below it in the basement.
When you consider each of the tasks involved in designing and maintaining a wireless LAN, it can really become a puzzle to solve. The cell size, transmit power, and channel assignment all have to be coordinated for each and every AP. Roaming also becomes an issue on a large scale, if mobile clients can move throughout an entire campus wireless network.
The good news is that Chapter 13, “Understanding RRM,” explains how to solve many of these puzzles automatically.