802.In-depth Day Three: 802.x performance

In the final part of his series on the 802.x WLAN protocols, Michael Martin explains the performance you can expect from these networks.

The terms wireless and performance do not really go together, even though we want them to. Wireless is convenient and provides a great deal of flexibility by expanding the computing environment beyond the desk and conference room. But you're not going to rip out your copper CAT5 and go completely wireless, at least not with today's technology. It is this "secondary network" mentality that contributes to wireless security problems. Connecting an access point (AP) is as easy as connecting a laptop to the wired network.

And that' were the problems begin. Getting back to performance, there is a large disparity in performance between wired and wireless data transmissions. This is due to the overhead needed to transmit the data using RF, instead of a physical medium. Working with theoretical maximums, TCP and UDP transactions over wired Ethernet utilize over 90% of the available bandwidth for data transport, as opposed to wireless Ethernet, where only a little more then 50% of the available bandwidth is usable for data transport.

Then there are the issues of RF coverage and node density. RF coverage is a problem that all network architects need to face. First, of all there are practical limitations to the amount of wireless channels available for use. To ensure that you do not have channel overlap, a proper RF survey needs to be performed. Walls, ceilings, and other structures can attenuate and block the RF signal this has both positive and negative impacts.

On the positive side, areas that are poor for RF use limit the effectiveness of war dialing and war driving and make it easier to use overlapping channels without any negative impact. The downside, of course, is that you will need a large AP density to provide good coverage. Here are some average performance expectations with a single AP in an open office environment, showing data rate to distance from the AP:

Coverage, however, is not everything. But it is where most architects focus their design efforts. When laying out BSS, the performance expectancy for data rate and node density needs to be defined. For maximum performance, you want to have the AP placed as close to the user population as possible and control the number of nodes supported on each AP, since performance degrades as the user density increases. A node target of 20 is a good balance between performance and user capacity. A simple way to estimate AP density is to figure out the minimum expected throughput performance, then divide that value into the supported throughput of the AP. For example, at a per user rate of 768Kbit/s, an 802.11b AP could support 7 to 8 nodes, where an 802.11a/ g could support 70. This is not exactly the most elegant forecasting method, but it provides a starting point.

Most APs will allow you to set throughput baseline, so if the node drops below the set rate the AP disassociates the node. The AP can also be configured to only associate certain users, or each AP can be configured on its own IP subnet with limited DHCP scopes. These are also great approaches to tightening up wireless security, but that's for another article.

Read more on Wireless networking