Community Articles

WLAN Performance

Kevin Davis, Senior Consultant, NetQoS Inc.

Many enterprises have installed 802.11b/g (54Mbps) radios throughout their networks expecting the performance of their new wireless connections to approach, if not meet, this theoretical bandwidth. Alas, the performance of these wireless networks invariably fails to meet expectations, and in some cases even becomes performance bottlenecks that require adding more resources for support and maintenance. What happened to all the bandwidth and performance promised by 802.11g, or for that matter, 802.11a?

Perhaps an illustration may help: Have you ever driven down an interstate highway while listening to an AM/FM radio station? Even before you leave town, you may notice that at certain points the radio signal fades for brief periods causing you to miss key words in a news report, sports event, or panel discussion. As you travel farther away from the radio station, at some point the signal becomes weak and inaudible and quickly fades out altogether.

The same phenomenon applies to wireless LAN (WLAN) connections. They are also prone to interference and attenuation (degradation of signal) from building materials and other electronic devices that may use similar frequencies (such as cordless phones, Bluetooth devices, or microwaves). The further away a wireless user gets from a wireless access point (WAP), the more likely they are to be affected by attenuation and interference. However, unlike a radio station’s audible signal, data that gets lost on a wireless network typically must be retransmitted, if only because most networked applications use Transmission Control Protocol (TCP) for transport. Retransmission of lost data also impacts performance, and reduces overall throughput for user application(s).

The following issues can significantly affect the performance of wireless networks:

* The shared nature of Wireless Access Points (WAP)

Protocol requirements that are backwards compatible to 802.11b

* WAP Implementation

The Shared Nature of Wireless Ethernet

Similar to wired Ethernet’s Carrier Sense Multiple Access/Collision Detection (CSMA/CD) protocol that permits workstations to share a common Ethernet hub, 802.11 wireless protocols support a protocol that permits the sharing of a common access point--namely, the Wireless Access Point (or WAP) that bridges between a wired network and a collection of wireless devices. This protocol is known as Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA). Just as with legacy Ethernet hubs, throughput decreases as the number of users on a WAP increases. In fact, the 54Mbps theoretical bandwidth is shared by all users associated to a particular WAP, just as they all share access to that same device.

When designing a wireless network, it's important to pay careful attention to the number of users in a given area that is covered by any single WAP. For example, a college auditorium that seats 300 students may need several WAPs in order to provide sufficient aggregate throughput to service all the computers likely to be in use at the same time in that kind of venue.

The same would be true for a large conference room that acts as the meeting place for top company executives or where multi-media will be accessed using wireless media. Enterprise class vendors and experienced consulting firms can suggest or supply best practices for user-to-WAP ratios to enhance throughput for wireless networks.

Protocol Requirements for Backward Compatibility

Unlike 802.11a, which operates at a frequency of 5.8GHz, 802.11g requires backward compatibility with 802.11b. Given that 802.11b and 802.11g radios cannot “see” each other (802.11b uses Direct Sequence Spread Spectrum (DSSS) while 802.11g uses Orthogonal Frequency Division Multiplexing (OFDM) transmission signals), some mechanism was needed to allow these two protocols to access the same media. The result was a return to the Request-To-Send/Clear-to-Send (RTS/CTS) mechanisms found in legacy technologies for occasions when 802.11b radios associate with 802.11g/b WAPs.

Two things are specified within the 802.11g protocol, which affect performance and throughput when a 802.11b radio is detected by a 802.11g/b WAP:

* The Request-To-Send/Clear-To-Send (RTS/CTS) protocol is enabled which increases the overhead of the WAP and decreasing transmission rates

* 802.11g radios will switch to using the slower 802.11b “backoff” times for the CSMA/CA protocol thereby decreasing throughput

The overall effect is a significant decrease in performance for all 802.11g clients associated with the 802.11g/b WAP; a decrease of over 70% can occur. This explains why one good rule of thumb for improving performance when implementing an 802.11g network is to remove all 802.11b radios!

WAP Implementation

When implementing a WLAN, three things can adversely affect performance and therefore need careful consideration and planning when deploying WAPs:

* Distance Between WAPs

* Selection of Channels

* Selection of Antennae

* Placement of WAPs

The distance between WAPs for network performance and complete area coverage depends on many factors. WAPs need to be close enough that users traveling between radios will not be dropped and signal strength will not adversely affect performance or availability. Here are the main factors that affect WAP placement:

* The frequency of transmission (5GHz radios need to be twice as close to each other as 2.4GHz radios because the 2x higher frequency travels only half the distance)
* Interference from building materials such as walls, structural beams, metal blinds, and elevators
* The number of users that may possibly associate to each WAP (it's necessary to put more WAPs where there are more users)

The channels for each WAP should be selected so adjacent radios will not interfere with each other. Typically, this produces a “honeycomb” pattern within the WLAN network where no two adjacent WAPs are configured to use the same channel.

High gain antennae should be selected for a WAP based on its physical location. Directional antennae should be used for WAPs connected to external walls (with the signal pointing inward), while omni-directional antennae should be used for WAPs installed on ceilings serving office cubes in open spaces. Also, external USB-based high gain antennae can be used on laptop computers to provide a performance boost.

WAPs should be placed at least seven feet above floor level and away from metal objects and other devices that may interfere with its transmissions. For example, do not place a WAP on a desk next to a row of metal file cabinets, or on a metal shelf in a wiring closet.

The selection of inter-WAP distance, radio channels, placement, and antenna types should all be part of a detailed site survey and design process conducted before installing any wireless network.

WLAN Visibility

So, how do you know you have your WLAN set up for optimal performance given the reality of a hostile environment and the nature of the 802.11 protocols?

Network instrumentation and software should be used to implement and monitor the performance of WLAN implementations. For example, end-to-end latency, utilization, throughput and retransmission reports can be used to optimize WAP placement (that is, empirical data for throughput achieved with WAPs x distance apart, or with y number of associated users). Where high values for retransmissions and low levels of throughput are observed, WAP configuration and placement should be reviewed and re-engineering as necessary.

802.11 Wireless Protocol Summary:

Loading...