The communications world is becoming increasingly dependent on wireless local-area networks (WLANs). New applications based on 802.11 standards are rapidly expanding into corporate, residential, Wireless-Internet-Service-Provider (WISP), and hot-spot applications. As with all new technology, the initial novelty eventually wears off. At that point, it becomes increasingly important to know how products will respond and perform in their intended environments. As the technology moves into the mainstream, a demand is created for more thorough and accurate testing.
The need already exists for throughput, latency, errored packet performance, and hundreds of other wired-based performance tests. Now, there also is a need to test the characteristics that are unique to wireless networks. These characteristics include: radio-signal impairments (fading, interference, Doppler effect, reflections, multiple paths, etc.); roaming hand-off delays (when a user passes from one access point to another); security; power management; and Media Access Control (MAC) -layer collision avoidance. Whether it is standard performance testing or unique WLAN network functionality, however, all WLAN testing must be done against repeatable, real-world radio-frequency (RF) characteristics. This is the only way to ensure valuable test results.
Compared to the testing of cable-based networks, the testing of wireless-network systems presents new challenges. For example, wireless-network performance can degrade significantly over time due to environmental changes in the signal broadcast area. WLAN service providers and IT managers must therefore perform network acceptance testing when they are turning over a newly installed network to their customers. To make sure network performance does not degrade below acceptable levels, they also must execute ongoing service-level testing.
This article is the first of a three-part series that will introduce WLAN testing requirements, test tools, and some of the significant issues facing design engineers, test engineers, and IT managers. Testing methodologies will be the focus of the second article. The third article will conclude with actual comparative test results provided by CENTAUR Laboratories at the University of Georgia.
As a standard WLAN RF signal radiates through its environment, it bounces off various obstructions like walls, file cabinets, and other reflective surfaces. As a result, multiple signal paths arrive at the receiver. Each of those paths is subject to independent delay and loss (FIG. 1).
When individual signal paths from the transmitter arrive at the receiver at different times, relative path delay occurs. The initial signal's net effect on the arrival time is to spread it out in time. In a digital system, this causes the received symbols to overlap. Called inter-symbol interference, this condition impairs the receiver's ability to successfully lock onto the signal and decode the desired waveform. This problem can cause a loss of data or the need for retransmissions, resulting in increased latency. The amount of relative path delay varies based on the environment (residential, business office, or commercial) and the application.
The receiver also is subject to relative path loss. Here, the individual signal paths arrive at the receiver with different absolute-power levels. These differences result from the power that is lost when the waveform reflects off of a physical obstruction.
The signal strength of an individual waveform will diminish according to the distance traveled. The loss of signal strength should follow the 1/d2 law, where d is the distance between the transmitter and the receiver. In actuality, the loss will be much greater (between 1/d3 to 1/d6). This is mainly due to the variety of reflective surfaces in WLAN environments.
The characteristics of these multiple paths are variable and fairly complex. It is thus desirable to have a standard way to simulate them in a repeatable fashion. If the tester is unable to simulate the same or similar multipath conditions over an entire test plan, it is very difficult to determine the actual performance of the device. It also becomes nearly impossible to conduct performance-comparison analysis from one product release to the next. For this reason, RF channel models (also known as fade models) are used.
A channel model is designed to simulate the signal characteristics of different frequencies. It also simulates different environments, such as offices, residences, or factories. The effects of each environment can be fairly complex. The channel or fade models attempt to generalize the complexities and establish a typical or average behavior for the channel in the environment in question.
A key parameter for any channel model is RMS delay spread. This is a measure of how the path delays are spread about the mean. Different environments will have different signal-reflection characteristics. As a result, they will have different RMS delay spreads. For example, the types of obstructions encountered in a warehouse and in an office complex will likely differ in delay and signal strength, resulting in different channel profiles. The channel profile specifies the number of independent paths, their relative path delay, and their signal strength. It is therefore extremely important to choose an appropriate channel model when testing for a specific type of environment (FIG. 2).
Over the last few years, a number of studies have been done on indoor fade models for WLAN applications. To date, several models are in use. But most of these were developed primarily for computer simulation. In some cases, it can be difficult to implement them in a real-world test environment with actual hardware. Here are some fade models that are currently in use for WLAN applications:
Exponentially Decaying Rayleigh Fade Model
This model, proposed by Naftali Chayat, is recommended in the 802.11 specification. It models a path delay and loss profile that follows a standard exponential decay. (Note that this proposed model became the baseline model for the comparison of modulation methods during the development of 802.11a.)
JTC'94 Indoor Fade Models
The Joint Technical Committee proposed these nine indoor fade models. They are based on different environments (residential, office, and commercial). By using different RMS delay spreads, the models achieve different channel profiles. These models also flaunt an advantage: A standards committee agreed upon them as models for wireless communications. Although these are accepted models, the RMS delay spread offering is limited. No models exist between 150 and 450 ns.
High Performance Radio Local Area Network (HIPERLAN/2) Fade Models
These models target a set of non-802.11 European WLAN standards. Though these indoor fade models were adopted for use in HIPERLAN/2 simulations, they can be used with 802.11a frequencies. Five environment-based models were developed, including an office and some large open spaces.
When choosing a Wireless Channel Emulator for the test lab, make sure these channel models are preconfigured and can be selected without the need to program each impairment parameter.
RF interference is unwanted radio-frequency energy. Specifically, this energy is emanating from other devices that are operating in the same WLAN frequency band. In addition to that interference, non-802.11 devices can generate considerable potential interference. Among such devices are microwave ovens, cordless phones, and Bluetooth devices. This is especially true in the already crowded 2.4-GHz frequency range.
To test a device's tolerance for RF interference under real-world conditions, combine a Wireless Channel Emulator with a noise and/or interference generator. A WLAN device that proves that it can operate in the presence of interference will result in happier customers.
To fully analyze the performance of a WLAN device in a repeatable, real-world environment, one needs a traffic generator. This generator must be capable of providing a consistent stream of traffic across the device under test. Traffic generators fall into two categories. Their categorization depends upon whether their processing power is focused on real-time, high-resolution error reporting or high-port-density load testing.
For proof-of-concept testing and the debugging of silicon chip or device designs, chip and equipment manufacturers often use traffic generators that provide real-time error reporting. To satisfy this testing requirement, the generator should be able to graph a variety of traffic conditions like latency, packet rate, and bandwidth utilization in real time. It also should be able to report errors at the packet, frame, byte, and bit level (FIG. 3).
Once the fundamental physical- and MAC-layer implementation is stable, it is important to see how well the design scales. Scalability testing can determine things like an AP's ability to handle multiple data streams per port or multiple station associations. (A station is a wireless client, such as a laptop with a wireless network-interface card.) Scalability is important for wireless-Internet-service and enterprise-equipment providers.
Specific test scenarios also should be conducted. For example, the usage patterns in a campus environment can cause large spikes in AP activity. The data traffic generator should be capable of simulating all of these conditions. It also should provide comprehensive results that can be easily compared and evaluated.
Intuitively, one would think that WLAN users aren't normally walking around using their laptops. Roaming hand-off delay should therefore not be an issue. A number of scenarios do exist, however, in which the time and efficiency that a wireless client disassociates and re-associates with an AP could be critical.
On the client side, applications are being developed for handheld devices in the medical field. While doctors are making their rounds in a hospital, they can enter their findings in a central database via a WLAN connection. In this example, roaming hand-off delays could impact the application's effectiveness.
In addition, a wireless client that is in the range of two access points can constantly switch its association between the two. The impact of re-association and the rate of association changes may vary. It is certain, however, that this ping-pong scenario can drastically impact performance.
Look at the issue of roaming hand-off delay from an access point's perspective. One must consider the characteristics of a campus or large enterprise environment. In these environments, APs must be able to handle huge spikes in roaming activity. Such spikes will occur when classes and meetings begin and end.
Rate adaptation is part of the roaming hand-off. When a client starts moving out of an AP's range, that AP tries to maintain the connection. It steps down the transmission speed, thereby adapting the rate. As the client continues to move away and the signal degrades further, the client has the option to look for another AP. When the client shifts from one AP to the next, the traffic flowing to it must shift its physical path from the first AP to the new one. This shift is not always instantaneous. The interruption can cause increased latency due to packet errors and retransmissions.
In the IEEE 802.11 specification, there are three classes of client mobility:
- No-Transition: Movement may or may not occur and the client never changes its AP association.
- Basic Service Set (BSS) Transition: A wireless client moves between multiple APs that are part of the same wireless infrastructure.
- Extended Service Set (ESS) Transition: A wireless client moves between multiple APs that are not part of the same wireless infrastructure.
The 802.11 specification states that service interruptions are likely during ESS transitions. It also implies that BSS transitions can be completed without interruption. An ESS transition requires the support of high-layer protocols like Mobile IP (MIP). Such protocols allow for uninterrupted mobility between sub-networks. In either case, the roaming-hand-off delay can be tested with the configuration shown in Figure 4.
To the roaming methodology being used, the roaming-hand-off-delay test scenario remains transparent. Regardless of how the wireless client changes its AP association, any traffic flows destined for that client must change their physical paths. Such an exchange will probably not be instantaneous. Indeed, it is likely to trigger various errors like lost, mis-inserted, duplicate, and out-of-sequence packets. It also can cause increased latency.
Wired Equivalent Privacy (WEP) is the encryption standard specified with 802.11. However, WEP encryption can be cracked with simple crypto-analytical attacks that are widely available. WEP may be acceptable for a home user. But it will not be popular with most corporate IT managers, WISPs, or hot-spot providers. The IEEE is working on a more comprehensive and robust security standard, dubbed 802.11i. The IEEE 802.11i working group should be submitting their specification for approval later this year.
This new standard will provide a security specification that is aligned with the National Institute of Standards and Technology (NIST). It will use the Advanced Encryption Standard (AES), the 802.1x authentication method, and various authentication algorithms. These algorithms include: Extensible Authen-tication Protocol-Transport Layer Security (EAP-TLS), Extensible Authen-tication Protocol-Subscriber Identity Module (EAP-SIM), and Protected Ex-tensible Authentication Protocol (PEAP).
Meanwhile, the industry has turned to other readily available security methods, including IPSec and Secure Sockets Layer (SSL). Temporal Key Integrity Protocol (TKIP) is the WEP replacement used in Wi-Fi Protected Access (WPA). Other methods include Extensible Authentication Protocol (EAP) algorithms, integral firewalls, and proprietary authentication algorithms. Two examples of proprietary algorithms are Funk's Extensible Authentication Protocol-Tunneled Transport Layer Security (EAP-TTLS) and Cisco's Lightweight & Efficient Application Protocol (LEAP).
With so many methods and implementations, different security measures will surely impact an AP or WLAN network's performance in widely varied ways. The networks should be tested for their ability to handle IP Virtual Private Network (VPN) tunnel creation. To test the VPN, send Hypertext Transfer Protocol (HTTP) or User Datagram Protocol (UDP) traffic over it. Test results should include a full complement of network test statistics, like packet loss and latency.
Due to the variety of security implementations, it is critical to match network design with security and throughput requirements. Test the chosen security methods thoroughly in the intended environment. This provides the information to assure WLAN users secure data. Plus, performance is not sacrificed.
One curious bi-product has been born out of the harnessing of wireless technology for communication networks: the "hidden-node problem." To visualize this problem, picture three people standing in a line some distance apart. They are trying to talk to each other. While the middle person can hear each person, the people at the ends of the lines cannot hear each other. If the person in the middle is not signaling the two people on the outside, they will eventually start talking at the same time. Because they can't hear each other, communications will break down.
To resolve the hidden-node problem, the 802.11 MAC layer uses Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA). Basically, it allows the transmitting station to request medium time by sending a Request To Send (RTS) frame. This frame contains the time at which it needs the medium. The AP will respond to the RTS with a Clear To Send (CTS) frame. This frame indicates the time when it will allow the station to use the medium. All stations associated with the AP will see the CTS frame. As a result, they will know not to transmit for that period of time.
By using packet fragmentation/defragmentation along with the RTS/CTS handshake mechanism, it is possible to optimize throughput by minimizing the potential for errors. From vendor to vendor and network application to network application, what varies is whether these two features are enabled and to what frame size they are applied. The objective of configuring RTS/CTS and fragmentation/defragmentation parameters is to give the WLAN the best throughput/error ratio performance.
The IEEE 802.11 standard also contains provisioning to support battery-powered wireless clients, such as laptops and handheld devices. This aspect will help conserve battery life by allowing the wireless client to reduce power and go into sleep mode. It is the responsibility of the AP to buffer data destined for the client. Although this process is well documented in the specification, the standards committee left the implementation of buffer management up to the implementer. This implies that power-management implementations will vary from one vendor's AP to another.
It is becoming common practice to leverage a wireless connection. In doing so, one can take advantage of the freedom of movement and increased productivity that it provides at work, on a campus, and in the home. However, the buying public can be quickly turned off by unrealized expectations—particularly when it comes to performance and quality of service. To prevent this from happening, test methodologies and comprehensive test plans must be defined. By simulating real-world conditions in the lab, the true performance of products and services can be ascertained and the enormous potential of WLAN can and will be realized.