For even relatively seasoned network administrators, designing an enterprise wireless LAN (WLAN) may be new ground. They may know how to plan for optimum capacity, for users and applications, on cabled networks but WLANs introduce a new factor - the unpredictable behaviour of wireless transmissions inside buildings. This leads to a trade-off between radio-frequency (RF) coverage and network capacity. Getting these two parameters right is the holy grail of designing WLANs.

The trick is to understand that network capacity has priority over network coverage - get capacity right and complete coverage will follow automatically. This may seem like putting the cart before horse, after all, its important that everyone in given area has access to the WLAN but in reality, the real challenge isn't to cover the largest area but how to deliver enough network bandwidth to meet the demands of business applications. Enterprise users are accustomed to high-speed, full-duplex 100 Mbps switched networks and expect similar performance from their wireless LAN connections.

So planning and designing a WLAN is key, if you're looking for success from day one.

Planning a WLAN is no different from planning other projects. There are a number of logical steps you need to follow in order to succeed.

System requirements
Because network capacity is not necessarily a given with WLANs, it is important to understand things like the number of users, their requirements, what applications they use and how the WLAN will interface with existing cabled networks. Once you've done this, you can prepare a specification for the project.

An important part of this is capacity planning. The number of simultaneous users that an access point can support depends mostly on the amount of data traffic at the time. Bandwidth is shared among users on a WLAN as with wired network connections. Network performance, as gauged by the number of simultaneous users, hinges on their combined computing activity.

So, what they use the network for is more important than how many of them there are. For example, web browsing is 'bursty' and incurs little network/radio traffic; office applications such as e-mail are quite light users of radio bandwidth. However, a large enterprise resource planning (ERP) or customer relationship management (CRM) application has much more interaction between the clients and servers and will cause much more network traffic.

On an 802.11b WLAN, each hardware access point has up to 11Mbit/s throughput. This would be adequate for:

  • 50 occasional users, light computer users that check the occasional e-mail
  • 25 mainstream users who get a lot of e-mail and download or upload moderately sized files
  • 10 to 20 power users who are constantly on the network and deal with large files

    Determining how much bandwidth each user will need is critical, as your estimates will define the user experience as well as the number of access points required. A good rule of thumb for a 54Mbit/s 802.11a/g network is to allow for a total of 4Mbit/s upstream/downstream per user, which delivers about the same user experience as being on a wired LAN. For an 802.11b network, a rule of thumb is to allow for 500Kbps each way or 1Mbps total, which delivers a user experience similar to a broadband Internet connection. To increase capacity, more access points may be added, which gives users more opportunity to enter a network.

    The next step is to produce a specification that will deliver a network capable of meeting the requirements. This includes technical elements such as developing the system architecture, identifying standards (e.g., 802.11b, 802.11g or 802.11a), specifying antenna types, and so on. It's important to bear in mind that, unlike traditional cabled network environments, when you lack network capacity, you can't 'throw bandwidth at the problem' by simply ordering larger circuits and installing larger boxes. With WLANs you can't 'buy more frequency' when network congestion becomes a problem. It may sound like a recipe for an admin nightmare but to address system capacity and plan for future user growth, network managers should consider deploying dual-band, 802.11b/g and 802.11a networks to ensure their wireless networks are not quickly overwhelmed by success.

    Site Survey
    The next vital chore is the site survey. Obtain a detailed floor-plan of the service area, with details of walls, windows, and so on. Go walkabout and look for anything that could affect transmission, things like metal shelving, lift shafts or plant and machinery. One chief technical officer wrote: 'When planning a wireless network you have to apply the "scream test:" Have someone stand on the other side of the wall and scream. If you cannot hear them through the wall, then the wireless signal probably won't pass through the wall'.

    Note where the users are located and position the access points accordingly, bearing in mind its range. Draw circles on the floor plan to indicate an AP's service area. Expect some overlap but ensure that channel assignments keep the AP's separated so as to not interfere. Don't forget: access points can overlap with other AP's on the floor above or below.

    To be thorough, you need to conduct radio-frequency (RF) tests, too. It's tricky to predict just what sort of service coverage you'll get from a particular deployment of a WLAN. Even if you're using omni-directional antennas, radio waves don't really travel the same distance in all directions. Walls, doors, elevator shafts and even people offer varying degrees of attenuation, which cause the RF radiation pattern to be irregular and unpredictable.

    Obstructions such as walls and furniture will cause signal reflections, a phenomenon known as multipath fading. This could create dead spots in the service area covered. While dead spots are difficult to predict they can be compensated for by physically moving the wireless equipment (usually dead spots are small and highly localized) or by re-orienting the device's antenna.

    Some WLAN vendors, including Cisco and Proxim, provide free RF site survey tools that identifies the associated access point, data rate, signal strength, and signal quality. You can load this software on a laptop or PocketPC and test the coverage of each preliminary access point location.

    RF interference is a further complication. Unwanted RF signals can unwittingly cause 802.11 devices to either fallback to a lower throughput or simply pause transmission until the interfering signal goes away, so it's important to know what's out there! 802.11b/g devices broadcast in the 2.4GHz waveband, a portion of the radio spectrum inhabited by microwave ovens, Bluetooth device other wireless LANs and exotic devices such as avideo senders.

    There is a long-standing myth that Wi-Fi interferes with DECT cordless phones, although in fact they operate on completely different frequencies. The idea that they might interfere appears to have been started by companies selling the failed Wi-Fi rival, Home RF, and persists after the death of HomRF, perhaps because some cordless phones in the US use the 2.4GHZ spectrum.

    Installation and Testing.
    Bear in mind, when placing access points there is a trade-off between distance and network throughput. Signal strength declines with distance and in order to maintain a reliable connection, an access point will throttle back throughput as signal strength declines. This graceful degradation or fallback is planned - for example, an 802.11b system will fallback from 11Mbps to 5.5Mbps as range increases or signal quality decreases. Subsequent fallbacks from 5.5Mbps to 2Mbps and 1Mbps are also supported. So access points should be close to areas where people are likely to want higher throughput.

    Installations should overlap the ranges of access points to ensure a seamless hand-off for roaming employees. However, adjacent access points have to use different radio channels to avoid interference. 802.11b and g access points have 11 channels but because of frequency overlap, there are, in fact, only three unique non-overlapping channels available. So you could place a maximum three 802.11b/g access points in a 'triangle' of adjacent areas, with each unit set to a different channel. By contrast, 802.11g kit uses the less congested 5GHz waveband. This gives is a shorter range but it does mean that it has up to 13 (14 in Japan!) non-overlapping channels.

    It's generally best to place the access points as high as possible, possibly above the ceiling tiles but remember to be practical. Warehouses have high ceilings; however, if possible avoid mounting them so high that you need ladders or scaffolding to get at your access points! You should also consider using Power over Ethernet (PoE) to run electricity to your access points. This avoids the time and costs of installing electrical outlets throughout parts of the premises that may not have power.

    For longer distances or environments with many physical barriers, you might need to employ WLAN bridges. A bridge will require a power outlet, but doesn't need to be wired into the network. The bridge simply takes the signals from the air and rebroadcasts them to their intended destination.

    After completing the installation, be sure to test for proper operation and coverage using 802.11 analyzers such as those from Airmagnet or Wildpackets. You may have to adjust access point locations or antennae and so on. Once optimal coverage has been achieved, access points can be connected to the LAN.

    Conclusion: Actual use
    This covers the practicalities of installing a wireless LAN. However, you must also make sure you consider security -- both in terms of securing the data that is carried on the LAN, and also access control, so the wireless LAN does not provide an unprotected gateway onto your LAN.

    These issues are all well understood. TechWorld will continue to deliver best-practice ideas over the coming weeks.