Planning Access-Point Placement

At this point, most users perceive that wired networks just work. In terms of product maturity and predictability, wireless networks are a step backwards. Protocols still behave somewhat unpredictably, and the relatively young protocols will require you to roll up your sleeves. When planning a network, you need to put in a great deal of work up front. Because wireless networks make user-location transparent to resources, the network itself must be aware of user location.

Once you have an understanding of the requirements and which 802.11 physical layers to use, it is time to figure out where to put the access points themselves. Depending on requirements and budget, figuring out placement may be a quick and dirty project that takes a few hours, or it may involve several visits and a great deal of time and money. This process is often called a site survey because some of the work takes place at the location where the network is being installed, but a new wave of tools is making it possible to do much of the access point placement work with computational models.

Several options exist for determining where to put APs. Vendors may provide site surveys to early adopters who agree to be reference accounts, although the time to be an early adopter of wireless faded long before this book hit the press. Value-added resellers may also have the skills to perform detailed site surveys; resellers may sell site survey consulting services or use site surveys as a way of coming up with a wireless LAN deployment bid. Some companies that specialize in technical education also offer classes on performing site surveys.

The Building

One of the major constraints on where access points are placed is how the building is constructed. Walls, doors, and windows get in the way of radio signals. It is a great help to get blueprints or floor plans and take a tour of the installation site as early as possible. Occasionally, you may get the opportunity to work on a building as it is being constructed. Seize any opportunity for a walkthrough because you will get a much better view of the interior structure before the walls are up. The main drawback to working with a building under construction is that experiments are not possible until the building is finished.

Once you obtain the floor plans, note where coverage must be provided. If the architectural plans are complete and include wiring information, they should also note the location of nearby power and electrical drops. You may also be able to make a preliminary identification of structures that may be problematic, such as ventilation ducts or reinforced concrete walls.

Whether or not you are using an electronic tool to automate the planning process, you should take a walk through the facility. Look for any changes that are not reflected on the blueprints. Verify the type of construction. Unless the building is old and historic, the walls are likely to be made of sheetrock hung from studs, although you can verify this by tapping on the walls. Ask whether there are any firewalls (of the physical sort, not the network sort), since a firewall is likely to be a significant barrier to RF. Structural, or load-bearing, walls are also likely to have reinforcements that pose a significant obstacle to radio waves. If you must conceal APs in certain areas, investigate these areas thoroughly to determine how to hide them, as well as any potential impact to network operations. High ceilings or other difficult mounting should also be investigated. During the walk-through, you are likely to receive a number of strange looks; just keep moving. Thorough walk-throughs will result in a large number of strange looks. Consider it a measure of the quality of the walk-through.

Based on the first walk-through, note any relevant environmental factors. Most importantly, you can correct the blueprints based on any changes made to the structure since the blueprints were drawn. Many minor changes will not be reflected on the blueprints, especially if the building is old. Also note any potential sources of interference. The 2.4-GHz ISM band is unlicensed, so many types of devices using the band can be deployed without central coordination. Newer cordless phones operate in the 2.4 GHz band, as well as Bluetooth-based devices and a number of other unlicensed radio devices. If you anticipate a large amount of interference, testing tools called spectrum analyzers can identify the amount of radiation in the wireless LAN frequency band. In practice, spectrum analyzers will not be necessary much of the time unless there is a particularly stubborn source of interference. One handheld analyzer, the Berkeley Varitronics YellowJacket, has a built-in spectrum analyzer for the 2.4 GHz ISM band, and can be useful in tracking down non-802.11 interference.

If your organization does RF testing, it may be necessary to shield any labs where testing is done to avoid interference with the wireless LAN. As a rule of thumb, keep access points at least 25 feet away from any strong interference sources.

Constraints on AP placement

Practically speaking, keeping the APs as high as possible means mounting them around the ceiling level. In many office buildings, the ceiling is a "dropped" or suspended ceiling. APs may be attached to the ceiling mounting bars, or possibly placed above the ceiling tiles. In a pinch, it is possible to mount APs at the top of cubicles. Check with your manufacturer to see what the radiation pattern looks like. If an AP is designed to send radio waves down from the ceiling, it may not work as well when mounted on top of cubicle partitions.

It is relatively common to make adjustments to the AP locations based on physical constraints. One of the primary restrictions is that APs typically connect to the network through an Ethernet cable.[*] Each AP must be within a 100-meter cable length from the nearest wiring closet. With some flexibility in the specifications, it is occasionally possible to run cables slightly longer than the official limit, although it is never a good idea to rely on it. Depending on the exact configuration of patch panels, risers, and the cabling in the ceiling, the cable run to an AP location may be much larger than it appears.

[*] A few products can use a secondary radio as a "mesh" backhaul, and I expect this option to be more common as time passes.

Electrical power is often a constraint on the AP locations as well. Early APs required standard electrical outlets, but very few organizations had electrical outlets in the ceiling. Installing electrical power is generally quite expensive. With the development of 802.3af, most organizations are powering up their access points over the network cable. Power over Ethernet (PoE) has many advantages; however, the engineering difficulty of pushing 48 volts through 100 meters diminishes your ability to cheat on the cable length.

It is a worth considering installing new wire to support a wireless network, especially if the existing network jacks will not easily support the new APs. When existing wiring is used, it is often run near the baseboards or otherwise less than ideal for use with high-mounted network devices. Running new cable has several advantages. It can be run directly to the area above the ceiling where the APs will be mounted, which removes the need for ugly cable runs up network poles. Older cable systems may have been installed before Category 5 wiring was standard, or modifications may have degraded its capabilites. New cable installations also add a degree of flexibility. In recognition of the need to move APs, some new cable installations will terminate the cable run above the ceiling with a network jack, allowing the use of patch cables to move an AP within short distances.

Depending on the layout of the components above the ceiling, it may be hard to run cabling near ventilation ducts, or hard to mount an AP near a duct. Usually, a relocation of the AP to the next tile over is sufficient and does not radically disturb the coverage of the AP. Likewise, keeping APs away from light fixtures, as well as the electrical conduits to those fixtures, is also important.

A final consideration relates to physical security. Some organizations may feel the need to secure APs against theft. Sometimes, mounting APs above the ceiling may be a sufficient theft-prevention measure because the AP is hidden from view. Many APs now include security slots for physical attachment to an less movable object.[*]

[*] See http://www.kensington.com/html/1356.html for more information on the Kensington Security Slot, which is commonly used for this purpose.

Buildings in progress

With the increasing use of wireless LANs, many are being designed during building construction. Working with a building under construction is both a great deal of fun and a sanity-defying challenge.

One of the first problems is a dependence on the other schedules for the building. Fortunately, it is possible to get an idea of the number of APs necessary by using an modeling tool. Architectural drawings are complete, and generally readily available to the network team even before ground is broken. Start by estimating the various types of material and developing a model.

Walk through each area of the building as it takes shape. Depending on the schedule, it may take shape in several stages or all at once. Gaining access for a walk-through during construction often requires the permission of the contractor, a hard hat, and a liability waver. In each area of the building, try to determine if the guesses made during the modeling stage were correct. Was the radio loss factor assigned to the walls accurate, or does the model need refining? As areas are finished, it makes sense to spot-check predictions from the modeling tool with a physical test using an AP in its final location.

One of the biggest challenges in working on a building under construction is that the environment is constantly changing. In rare circumstances, the building materials may change due to supply shortages or even last-minute changes driven by aesthetic considerations. Keep the design flexible, and build in a margin for error.

The Preliminary Plan

Determining the number of APs and developing a set of preliminary locations is the first planning milestone. A few years ago, the best way to come up with preliminary locations was to estimate the number of very expensive APs, and place them in open areas that were centrally located. Now that APs are much cheaper, there is no longer major pressure to reduce the AP count as far as possible.

To get a very rough ballpark estimate of the number of APs needed, I use two rules of thumb. The first is area-based. At maximum power in a typical open office environment, an AP can provide service to an area of 3,000 to 5,000 square feet (275 to 450 square meters). Simply take the area to be covered, and divide by the AP footprint. Use the higher number if there are lots of open areas and relatively few obstructions; use the lower number if there is a preponderance of closed-wall offices with doors, or significant interior structure to the building. In addition to the area-based number, calculate a user density-based number. Divide the number of users in the organization by 20 to 50. Use a number at the low end if wireless networks are prevalent and widely used, and a number at the high end if wireless networks are new and experimental. By taking the larger of the two numbers, I have a very rough estimate for the number of APs required for an installation.

Knowing roughly how many APs are needed does not provide any information on where they should go. Turning the extremely rough estimate into a preliminary plan requires a bit more work. With enough experience, the AP locations are fairly straightforward. Structural walls or fire-containment walls tend to block a signal, often completely. The central core of a multifloor building is usually the main load-bearing structure, and is often made of reinforced concrete. With attenuation of 10 dB to 20 dB per foot, it is best to consider it an RF shield. Most APs are still usable at greater than the minimum speed after the signal has passed through two to three average-sized offices, depending on the amount of straight-line propagation. Generally speaking, locate APs in open areas with as few obstructions as possible. It always helps to keep them accessible for any needed service, which generally means that areas above cubicles and hallways are preferred.

Developing the preliminary plan has, up to this point, been largely a manual process that requires high skill and extensive experience. In the past few years, modeling tools have emerged that can automate the preliminary planning stage. These tools take an architectural drawing of the building and create a mathematical model of radio propagation. When the network designer changes the location of an AP in the model, the tool recalulates its coverage immediately. Electronic tools can be a valuable asset to reduce the physical validation by deriving a reasonable starting set of AP locations. They are especially valuable when the building does not yet exist because it is still under construction. Depending on the tool, you may be able to use the architectural drawing itself, which will typically be a Computer-Aided Design (CAD) file; some tools will only accept simple graphics files like GIFs or JPEGs. Obtain the physical layout. In most organizations, the facilities department has architectural drawings, or can work with the landlord or building owner to obtain them. In new construction projects, an architect can usually supply them.

As helpful as modeling tools are, it is important not to use them in a vacuum. Radio propagation is very complicated, especially indoors at microwave frequencies. Radio waves may reflect in strange ways off different types of materials, and a few inches here or there can make all the difference in the world to the resulting coverage. Modeling tools also depend on accurate knowledge of the construction of a building, which is not always available to the network staff. Modeling tools are no substitute for actual experimentation.

No matter what technique you employ, use the physical plans to derive a preliminary plan. Detailed radio channel use planning is not yet necessary. If you are using a modeling tool, have it suggest a radio plan, but expect to make some changes after installation. The preliminary plan is, well, preliminary. Expect it to change. The purpose of this stage is not to get everything right, but to get a starting point to work from. Table 23-5 is based on the best-case coverage radius from a typical omnidirectional antenna. If you want to incorporate higher than the minimum data rate, you will probably need smaller radii; the best approach is to use one of the electronic tools discussed briefly in this section.

The preliminary report

The preliminary plan can be used as the basis for further activity. It may be used to provide an estimate for the cost of deploying the wireless LAN in the given area. It may also be used to hire a cabling contractor to perform the physical wiring. The plan itself may include:

1. A brief summary of requirements.
2. Estimated coverage areas based on the site survey measurements. This may be divided into areas with good coverage, marginal coverage, and weak coverage. Reports based on mathematical models may note projected contours for different operational rates.
3. A description of the locations of all access points, along with their configuration. Software tools for automated layout may be able to provide detailed location and configuration information based on the floor maps, such as:
   1. The AP's operating channel.
   2. Approximate coverage area, possibly shown as coverage contours at different speeds.
   3. IP configuration, if required. "Lightweight" APs may not have an IP address.
   4. Antenna type and configuration, including direction for directional antennas.
   5. Any other vendor-specific information that may be useful. Some organizations track devices based on MAC address, so it would be useful to include that in the report.

Radio Resource Management and Channel Layout

Figure 23-4 makes the channel layout appear deceptively simple. When building a network indoors, signal propagation is much more complex. Channel overlap is far more likely indoors because of the need to have higher power directed towards obstacles, which may result in higher than desired power in other directions.

Preliminary plans should include a basic channel map. It is almost certain that adjustments will need to be made to the channel layout. Tuning may be done manually by searching out overlapping channels with handheld tools, or automatically by access points that attempt to find the least busy channel. With only three channels in the 2.4 GHz band, changes may "ripple" through a network and take some time to converge on a final solution.

Refining and Testing the Plan

Depending on the need for accuracy, as well as budgetary constraints, the process of refining the plan may vary a great deal. In small or budget-conscious installations, it may suffice to start with the preliminary plan, turn it on, and see what develops. More methodical deployments may install all or part of the preliminary plan and perform extensive tests to verify it meets requirements. If the preliminary plan is accurate, there may not be much modification required. The major goal of the testing is to discover any unforeseen interference or bad spots and redesign the network accordingly. In most cases, interference problems can usually be repaired by relocating an access point. Adjustments to AP locations are typically not large. Failing that, a different antenna or another AP is usually the answer. In rare cases, multiple design and test phases may be used, although it is quite unusual.

When checking the plan, duplicate the user experience as much as possible. Obstacles between wireless LAN users and access points decrease radio strength, so make an effort to replicate exactly the installation during the site survey. Antennas should be installed for the test exactly as they would be installed on a completed network. If office dwellers are part of the user base, make sure that adequate coverage is obtained in offices when the door is closed. Even more important, close any metal blinds, because metal is the most effective radio screen.

Signal measurements should be identical to the expected use of the network users, with one exception. Most site survey tools attempt to determine the signal quality at a single spatial point throughout a sequence of several points in time, and thus it is important to keep the laptop in one location as the measurement is carried out. Taking large numbers of measurements is important because users will move with untethered laptops, and also because the multipath fading effects may lead to pronounced signal quality differences even between nearby locations.

Extremely thorough organizations may want to test with multiple client devices. Wireless LAN behavior can vary a great deal depending on implementation, and client behavior often varies even with the exact same software. If you need detailed knowledge of how the system will behave, it may be worthwhile to gather several systems with identical software configurations and run them through the same tests at the same time.

The final test report should result in detailed knowledge of the actual coverage of the APs in their final locations. Coverage may be reported as an area, although it is often more useful to report on the area over which a certain transmission rate is reliably obtained. In some cases, it may also be useful to report on performance characteristics, especially if application mix is well-enough known to be characterized.

If the building is under construction, validation may proceed in stages as the building is completed. If the entire building is constructed at the same time, you may want to take basic measurements early in the process to determine if major changes to the radio model are required, with more time-consuming and accurate validation tests at the end.

Validation and test tools

In the past, getting an idea of the coverage of an AP and coming up with an AP layout was time-consuming because it required repeated measurements of signal quality as an AP was placed in a trial location. With the development of automatic layout tools, whether based on modeling or self-tuning capabilities, the validation phase may not require the same extensive set of tools. In-depth analysis of the radio link performance is generally only required when trouble with a test system is discovered.

The most common signal quality measurements are the packet (or, more properly, frame) error rate (PER) and received signal strength (RSSI). Frame error rates need to be kept as low as possible. In the past, an 8% target generally guaranteed acceptable performance. More densely deployed networks should easily be able to achieve 5% or lower. Sophisticated infrastructure devices can often measure the per-client frame error rate directly, as well as the RSSI and signal-to-noise ratio. The RSSI and signal-to-noise ratio are important for achieving higher data rates. To transmit an intelligible frame at a given data rate, there is a given signal-to-noise threshold.

A few tools report multipath time dispersion, which measures the degree to which a signal is spread out in time by path differences. Higher delay spreads make the correlation of the signals more difficult. With a high delay spread, devices need to accept either a higher error rate or fall back to a more conservative coding method. Either way, throughput goes down. The higher the delay spread, the more throughput suffers. Measuring multipath dispersion is generally not important, although it is valuable to have a tool that gathers the data for persistent multipath problems.

When 802.11b was the wireless network of choice, a large number of handheld tools were developed for portable computing platforms such as the Compaq iPAQ. With the increasing adoption of 802.11a and 802.11g, though, handheld devices are falling out of favor and being replaced by Tablet PCs. Most handheld devices have relatively slow external interfaces that are unable to cope with the higher data rates from newer standards. The iPAQ, for example, has an interface for 16-bit PC Cards that run at 8 MHz. While such slow speeds are sufficient for an 11 Mbps 802.11b interface, the higher speed of 802.11a and 802.11g imposes a practical requirement for CardBus. Tablet PCs have a second advantage in the validation phase as well. Many of the planning tools that you might use to develop a preliminary plan include the ability to run a validation client that collects physical measurements and can tie the validation measurements back to the predictions calculated earlier.

Particularly stubborn interference may require the use of a spectrum analyzer to locate the source of interference from a non-802.11 network. Devices that can scan a wide frequency band to locate transmissions are not cheap. Expect to pay several thousand dollars, or you can hire a consultant and rent one. As an alternative, it may be possible to use an ISM band-only spectrum analyzer to track interference to its source. In any case, a spectrum analyzer is the tool of last resort, necessary for only the most stubborn problems.

RF fingerprint collection

Some wireless LAN systems require the collection of RF "fingerprints," as discussed in the previous chapter. Fingerprint collection can only proceed once the APs are in their final locations. At that point, test devices can be positioned in common locations so the system can collect the result fingerprints.

Preparing the Final Report

During the planning and testing cycle, you should have prepared preliminary documents that show AP installation locations. When the final test has completed, the network should be documented. If the preliminary plans are in electronic form, modify them to reflect the results of testing, and incorporate the changes into a final report.

In addition to access-point placement, many customers appreciate an estimate of the work necessary to install drivers onto any affected laptops. The scope of this item depends a great deal on the sophistication of the customer's management tools. For many, it will be sufficient to include a copy of the driver installation instructions as an appendix to the report. Some clients may require low-level details on the driver installation so that the driver installation can be completely automated down to any necessary registry changes on Windows systems.