When wireless LANs (WLANs) first became available in the early 1990s, the primary applications were wireless bar code solutions for needs like inventory control and retail price
Today enterprises are deploying wireless LANs for larger numbers of users with needs for corporate applications that involve email, Web browsing, voice and access to various server-based databases. The need for higher data rates and techniques to improve WLAN performance is becoming crucial to support these types of applications. To get that extra performance for your clients, you have a lot to consider.
Choose the right 802.11 physical layer. An important element that impacts WLAN performance is the selection of the appropriate physical (PHY) layer (e.g., 802.11a, 802.11b, 802.11g or 802.11n). 802.11n hasn't been officially ratified, but it offers the highest capacity and is backward-compatible with earlier standards, such as 802.11g and 802.11a. An issue with 802.11b and 802.11g is that they are limited to three nonoverlapping radio frequency channels, making them somewhat limited in capacity. For maximum WLAN performance, choose 802.11n, but ensure that the vendor can provide easy future upgrades to the ratified version of the standard.
Properly set access point channels. The 802.11b and 802.11g standards define 14 channels (11 in the U.S.) that overlap considerably, leaving only three channels that don't overlap with each other. For access points that are within range of each other, set them to different channels (e.g., 1, 6 and 11) in order to avoid inter-access point interference. Your client can also take advantage of the automatic channel selection features that some access points offer. I often see companies setting their access points all to the same channel. The problem with this is that sometimes roaming will not work as users move about the facility, and the transmission of a single access point blocks all others that are within range. As a result, WLAN performance degrades significantly. This is not an issue with 802.11a because the 802.11a standard defines separate, nonoverlapping channels. 802.11n also offers an extended set of channels to operate.
Provide adequate RF coverage. If access points are too far apart, then some users will be connecting to the wireless LAN at something less than the maximum data rate. For example, users close to an 802.11g access point may be operating at 54 Mbps, whereas a user at a greater distance may only have 6 Mbps capability. In order to maximize WLAN performance, ensure that your client's RF coverage is spread out as much as possible in all user areas, especially the locations where the bulk of users reside. The completion of an effective RF site survey will aid tremendously with this exercise. The proper setting of transmit power and selection of antennas will also help in positioning your client's access points for optimum WLAN performance.
Avoid RF interference. Cordless phones and other nearby wireless LANs can offer significant interfering signals that degrade the operation of an 802.11b and 802.11g wireless LAN. These external sources of RF energy in the 2.4 GHz band periodically block users and access points from accessing the shared air medium. As a result, the performance of your client's WLAN will suffer when RF interference is present. You should strive to minimize sources of RF interference and possibly set the access point channels to avoid the interfering signals. Again, an RF site survey will help you discover interference problems before designing and installing the wireless LAN. If it's not possible to reduce potential interference to an acceptable level, then consider deploying 5 GHz, 802.11a or 802.11n networks.
Consider RTS/CTS. The optional request to send/clear the request-to-send/clear-to-send (RTS/CTS) protocol of the 802.11 standard requires a particular station to refrain from sending a data frame until the station completes an RTS/CTS handshake with another station, such as an access point. RTS/CTS reduces collisions associated with hidden nodes and may improve WLAN performance for your clients. Collisions can occur when hidden nodes blindly transmit when another station (blocked by some obstruction or significant range) is already transmitting. This causes a collision and results in each station needing to retransmit their frames, doomed again by a possible collision due to the hidden node scenario. The outcome is lower throughput. If you suspect hidden nodes are causing collisions/retransmissions on your client's network, then try setting the RTS/CTS threshold lower through a trial and error process while checking the impacts on throughput.
Fragmentation. An 802.11 station can use the optional fragmentation protocol to divide 802.11 data frames into smaller pieces (fragments) that are sent separately to the destination. Each fragment consists of a MAC Layer header, FCS (frame check sequence) and a fragment number indicating its ordered position within the frame. With thresholds properly set, fragmentation can reduce the amount of data that needs retransmission. RF interference often causes only a small number of bit errors to occur. Instead of resending the entire data frame, the station implementing fragmentation only needs to retransmit the fragment containing the bit errors. The key to making fragmentation improve throughput is to set your client's thresholds properly. A threshold set too low will result in smaller fragments (making retransmissions efficient), but the greater number of fragments requires substantial overhead because of the additional headers and checksums. As with RTS/CTS, use a trial-and-error process to set the threshold while keeping an eye on consequential throughput. If there is no appreciable RF interference, then it's best to deactivate fragmentation.
About the author
Jim Geier is principal consultant of Wireless-Nets Ltd. and assists companies with the design, implementation and testing of wireless LANs. Jim is author of over a dozen books, including Deploying Voice over Wireless LANs (Cisco Press) and Implementing 802.1X Security Solutions (Wiley).
This was first published in April 2008