How can you predict large-scale Wi-Fi performance?

When you buy a wireless LAN system, how do you know how well it will perform day in, day out in terms of end-user experience? I’m talking about in real-world, variable radio-frequency (RF) conditions. And as your network grows.

Currently, there’s no reliable way to tell. You can read vendor claims on product data sheets and see test measurement numbers that have been derived under controlled conditions. But, unlike wired networks that carry a certain amount of predictability by the “protected” nature of physical cabling, wireless networks perform better or worse depending on the idiosyncrasies of the RF environment and each vendor’s 802.11 implementation.

Also, WLANs’ shared medium – unlike wired Ethernet’s switched medium – throws a wrench into knowing how many clients a given access point (AP) can support and how much bandwidth will be available to a client at any given time.

For these reasons, the industry could use some formal benchmark tests that simulate large numbers of clients and APs under various environmental scenarios. To that end, an effort called 802.11.2 has been quietly under way for some time.

Undertaken by 802.11 Task Group T, this initiative aims to provide a set of measurement methods, performance metrics and test recommendations so that 802.11 systems can be consistently compared on wireless-specific metrics, such as rate vs. range under different RF signal attenuation conditions, roaming delay times, and how long it takes for a client to authenticate to an AP or WLAN controller. These measurements are in addition to traditional “wired” metrics such as throughput, packet loss, latency and jitter.

A pivotal character driving 802.11.2 test recommendations is the venerable Scott Bradner, senior technical consultant at Harvard University and a Network World columnist. Bradner has a track record: He was a primary force behind the creation of standardized benchmark tests for wired LANs 10 years ago.

The Task Group T’s recommended best practices for benchmark testing, now in revision 7, are expected in early 2008. More on this situation – and some expected related Interop activity this week – next time.

Finally, a request: If you can find a few minutes to take the Webtorials WLAN State-of-the-Market survey, I’d really appreciate it. For the third consecutive year, I’m collecting information from those deploying or considering deploying WLANs to track usage trends. The results will be compiled into a findings report, offered free of charge, in June. You can access last year’s report here (abstract without registration). Thanks in advance.