Craig Mathias: Site surveys are no longer needed
In reviewing proposals for enterprise wireless LAN installations, I'm amazed at how often I find a hefty sum quoted for a site survey. Much to the dismay of the installer, I almost always recommend deleting this item, because it's usually a waste of time. Site surveys have no relation to real-world performance. They're relics of a time when access points were expensive, so it made sense to optimise for coverage. But a site survey provides a moment-in-time snapshot of the radio environment; it doesn't consider the number of users, their traffic patterns and data loads, overall throughput requirements, interference or network changes.

Given that today's access points are much less expensive and still declining in price, while installation labor costs are rising, site surveys in effect convert inexpensive access points into an expensive, labor-intensive activity that ultimately yields sub-optimal results. Why do we still assume a philosophy of scarcity, optimising for the minimum number of access points required for basic coverage, instead of providing the abundance of capacity that leads to happier users, more productivity and the bandwidth that ultimately will be needed to support VoIP over Wi-Fi?

The philosophy I recommend is not unlike that used to develop software. Attempting to optimise the performance of an application before it's written is usually futile. It's better to get something working, measure where the bottlenecks are and then optimise those modules that need it. The same holds true for WLAN installations: It's best to deploy - taking into consideration user locations, traffic patterns and throughput requirements - and then use additional fill-in access points to compensate for any errors in initial assumptions.

Proponents of site surveys will counter that it's preferable to understand the radio propagation characteristics of a site and any potential sources of interference, upfront. However, site surveys cannot address these concerns - radio propagation is so nonlinear that no great truths usually emerge from the survey, and interference is best addressed via a spectrum analyser, not a WLAN.

Finally, two important technologies are further nailing the site survey coffin shut. One is radio frequency spectrum management (RFSM) tools, available from many WLAN vendors. These automatically configure (and re-configure) parameters such as radio channel and transmit power dynamically, and often can point out holes in coverage and under-capacity in a given location. Second, as we enter the era of dense deployments, as opposed to the sparse deployments resulting from the cult of scarcity that site surveys reinforced, we can now attach access points directly to wiring in the walls and essentially over-provision an installation while minimising installation costs [Aruba's recently launched Grid Point architecture is an example - Editor].

Density, coupled with centralised management and RFSM tools, is the key to successful WLAN installations as the WLAN evolves to be the default network connection for many companies. So skip the site survey and buy a few more access points. You'll be glad you did.

Mathias is a principal at Farpoint Group, a consultancy specialising in wireless communications and mobile computing.

Michael Hong: Site surveys are still needed
Site surveys are still a useful tool for risk management in enterprise wireless LAN deployments where business-critical applications such as wireless VoIP require carefully planned coverage and capacity. In the past, site surveys were used to determine where to deploy access points to provide the required coverage and capacity with the fewest number of access points. Site surveys determined access points' transmit power and channel settings to minimise interference with other access points. But two recent developments - intelligent access points and low-cost thin access points - have altered the goals for site surveys in enterprise WLANs.

Intelligent access points can now monitor their coverage area and automatically optimise their transmit power and channel settings, a welcome improvement over site surveys. Thin access points have reduced the coverage and capacity cost. Instead of using site surveys to optimise the location and number of access points, thin access points now can be deployed in convenient and easily accessible locations that are less than optimal for coverage, but because the thin access points cost less, additional access points can be deployed to compensate.

However, these developments don't make site surveys obsolete. Until there is an access point on every desk, at every cubicle and in every room to provide complete coverage and capacity requirements, site surveys are still a valuable tool, especially in enterprise WLAN environments.

Today "an access point everywhere and anywhere" is not a realistic WLAN deployment option. Access points can use existing Ethernet connections, but it is unlikely that there will be enough of these connections to deploy access points everywhere and anywhere. If there are not enough existing Ethernet connections available, new Ethernet connections will be required, which can be expensive and disruptive to install. And the cost of purchasing an access point is not limited to its price. Thin access points might be less expensive than thick access points, but they do not include the price of the required wireless switch or appliance and any access point license fees. For these reasons, decisions will need to be made on where and where not to deploy access points, increasing the risk that the coverage or capacity will be inadequate.

How significant is this risk? That depends on how critical the coverage and capacity is for your WLAN applications. In an office environment, inadequate coverage or capacity might result in a simple help desk request that can be corrected by deploying an additional access point. However, in a hospital, this might result in the inability to contact a nurse or doctor on a wireless VoIP telephone during an emergency.

Site surveys can assist in managing this risk by confirming that the WLAN provides the required coverage and capacity. As more organisations embrace WLAN applications such as VoIP, the need to manage this risk becomes increasingly important.

Who's right? Would you be happy never to do another site survey, or is the technology not yet good enough to do without it? Join the discussion in our Forum.

Hong is WLAN product manager at Foundry Networks. He can be reached at [email protected]