February 17, 2004
When deploying a wireless LAN, you need to know how many users the access points will support. This helps in determining whether the WLAN will bear a specific application. For example, the WLAN may need to handle twenty users accessing the Internet from an airport concourse.
Or, hundreds of users may need access to the WLAN from a convention center. Because of the shared medium access protocol of 802.11, however, capacity determination becomes a daunting task.
You can’t just rely on dividing up the data rate as a measure of capacity. It’s tempting to say that an 11Mbps data rate with 802.11b will provide ten users with 1.1Mbps of throughput each. Keep in mind, though, that the data rate is the speed at which the bits in the data frame are sent.
There’re delays between frames due to sensing time, transmission of acknowledgements, presence of RF interference, and other users taking turns sending data frames over the shared air medium. As a result, the actual throughput of a WLAN is much lower than you’d expect.
In fact, a lone user on the WLAN will likely experience 6Mbps throughput with an 11Mbps association with the access point. As the number of active users increases, collisions cause the throughput per user to decrease exponentially.
The point here is that it’s very difficult to determine the capacity of a WLAN using simple arithmetic. Instead, you need fancy mathematics that takes into consideration that access to the medium is a random event.
Rather than make this a graduate engineering class, let’s look at some practical methods to assess the capacity of a WLAN.
The Brute Force Way
One way of sizing up a WLAN is to install the access points and see what happens. This approach doesn’t require much investment in time and brain power; however, you may never really know the maximum capacity until users start to complain.
For example, you may install a WLAN to enable users in an office to access Internet services. After first commissioning the system, only a handful of users may be active at any given time.
Within six months, the number of active users may increase to fifty, and you start to observe users complaining about slow performance. This is an indirect sign that access points are operating at capacity. Of course this result assumes that there are no other bottlenecks, such as slow servers or ISP connections.
Another rather manual method of sizing a WLAN is to gather up a bunch of people armed with wireless laptops and start hammering on the network. Just keep adding more and more active users until performance starts to decline to a level that’s not acceptable.
The maximum capacity could be the point at which users experience more than five seconds delay when loading a particular Web page.
This approach will provide accurate results, but it’s often not feasible. The difficulty is finding enough people and laptops to perform the test. In most cases, this type of testing is just too expensive. Certainly, there must be a better way!
Simulation Yields Upfront Predictions
Simulation programs such as OPNET run on a computer and imitate a WLAN under different situations. With simulation, you can artificially characterize WLAN components, such as access points, radio cards, and users.
A simulation calculates resulting throughput, which gives you a good idea of how many users can be active on the network.
A strong advantage of simulation is that it can be done before purchasing and installing the WLAN. With a simulation tool, you create a scenario that includes any number of virtual users with estimated utilization levels.
At a push of a button, simulation results tell you how the WLAN will behave. By adjusting the number of users, you can estimate how many users that access points can handle.
Simulation tools are rather costly, though, with prices in the tens of thousands of dollars. In addition, there’s a bit of a learning curve before someone can become proficient at developing the simulation models.
This can make the simulation option out of reach by some companies. Results are also only as good as the utilization level estimates.
Live Emulation Testing
If the WLAN is already in place, consider user emulation testing to find performance limits. Emulation is similar to having multiple people live on the network, except emulation only makes use of single piece of hardware. An emulator actually generates traffic similar to humans.
The only company that I know of offering this sort of equipment is Communication Machinery Corporation (CMC). Their EmulationEngine family of products enables the emulation of up to 64 concurrent users that individually associate with the access point and then send data frames. Traffic generators emulate a wide variety of traffic for various applications.
The $4,000 to $5,000 for CMC’s tools makes emulation a less expensive option than using simulation. Remember, however, that emulation requires that the WLAN be in place. Concentrating on post-installation testing only reduces the feasibility of using the capacity measurements as feedback for design changes.
You may have installed an 802.11b WLAN, for instance, and find that 802.11a was necessary to support performance requirements after running the emulation tests. For larger WLANs, I’d rather learn the need for such a design change using simulation and before purchasing and installing the access points.