Over the year, I have taken a good deal of heat for my argument that way too much time and money are being) devoted to the historically venerable eireless LAN site survey. My argument is that the return on investment is low today. I still maintain this position. However, recent discussions with a few industry players--particularly those working for firms that design and install Wi-Fi systems--have led me to the conclusion that blanket statements (and we analysts are positively legendary for these) are seldom a good idea, at least without a bit of context. So, let me explain and partially make amends for oversimplifying, but, regardless, reach the same conclusion.
To begin with a definition, a traditional (sometimes called “active”) site survey involves temporarily installing an access point (AP), and then walking around with a laptop or other mobile device running an application that records the signal strength (RSSI) of this AP at each location sampled. This is assumed to be a good way to get an idea of where APs might be placed in a production environment, if for no other reason than to identify potential dead spots and to gain some knowledge of RF propagation in a particular venue. Site surveys of this form were essential in the early days of access points, beginning around the mid-1990s, because APs were very expensive and application throughput demands were much lower than they are today. Optimizing for coverage seemed like a good idea, if not the only reasonable goal for that era.
Unfortunately, what we really have here a classic more-variables-than-equations kind of problem, and that goal that makes little sense today. Rather, we should be optimizing for capacity, not minimizing the number of APs required for the desired coverage. The goal should be to ensure that users have the reliability, availability and latency characteristics that optimize their personal productivity--people being much more expensive than APs. Yes, coverage still matters, but solving that challenge is much easier today: Just add more APs. In fact, whether a site survey is done or not, we always suggest budgeting about 10% of the total CapEx involved in any given installation for APs than might otherwise be indicated to allow for both the remediation of errors in positioning and to support growth in demand.
But how could we get the location of a given AP wrong if we do a site survey? Well, let me count the ways:
- First of all, radio propagation is so non-linear as to be quite literally unpredictable via any known analytical technique. In other words, it’s impossible to predict how or even if a radio wave will get from Point A to Point B under any given set of conditions and circumstances. And given that these conditions and circumstances change from moment to moment--and, of course, that a wide variety of client devices with wildly varying performance characteristics are likely to be in use in a given installation--the information from a site survey can quite literally be useless in determining the location of a given AP.
- Radio-wave propagation, and especial signal fading, are highly frequency-dependent. Given the broad range of frequencies applied in Wi-Fi--most notably at 2.4 GHz and 5 GHz, but soon to include 6 GHz and all of the channels that might accordingly be used--a comprehensive site survey would be economically infeasible. Add in real-time and automatic variability in transmit power, modulation, channel coding, and techniques like beamforming, MIMO, bandsteering, and multi-user MIMO (now bidirectional in IEEE 802.11ax, a.k.a Wi-Fi 6), and, well, forget it--a site survey can clearly generate wildly inaccurate results, and often does.
- Now let’s add in user mobility, which can create clusters of demand with no easing of the above-noted issues. It’s your money--spend it on a site survey if you wish--but it’s increasingly clear that site surveys are an artifact of time long past.
- And, remember, a site survey is just a single set of samples from a single point in time--a snapshot of data that’s often invalid the moment after it’s taken. Add people, rearrange the office and add real network loads, and you’ll quickly see what I mean here.
All of that being said, we do recommend regular perusal of management-console and assurance-system information, along with consideration of the application of tools based on artificial intelligence (AI) and machine learning (ML). These technologies will continually improve during the next few years, and can be on-station 24/7/365 with the ability to spot and even fix problems that few humans can match.
To be completely fair, there are times when a traditional site survey should be done--buildings with unusual construction, industrial settings, places (like hospitals) with thicker walls, and related challenges. (We once did a site survey in a power plant, for example). But for the open-office settings that are common today, we suggest a basic RF sweep for potential interference, a review of traffic requirements with an emphasis on location, and a review of Ethernet switch capacity and power. Then go ahead and deploy using that 10% overage budget to correct any errors--and add more capacity in the bargain.
See? We turn a one-time snapshot of questionable value and perhaps hefty real-dollar expense into a much better result. Couple this with analytics, up-to-date assurance tools, and solutions increasingly based on AI and ML, and, well, I’d call that a much more productive use of time and money. Much.