Wireless LANs: Complement to or replacement of network wiring?
Cabled systems fare well in cost, security,and throughput considerations.
Wireless local area networks (WLANs), specifically those based on 802.11x technologies, have already reached mass-market acceptance for home networks and public hot spots, and have started to achieve significant penetration within corporate LANs. A September 2002 study of 225 chief information officers, conducted by Morgan Stanley, showed that 32% had already deployed a wireless network somewhere in their organization. With the promise of improved productivity, mobility, and speeds up to 54 Mbits/sec, WLANs provide a tempting alternative to traditional unshielded twisted-pair (UTP) cabling systems with a fixed architecture approach.
This article presents analysis of 802.11x systems on three key fronts: throughput, security, and cost.
The current commercially viable wireless standards under the IEEE moniker 802.11 include 802.11a, 802.11b, and 802.11g. For each of these technologies, access point (AP) is the term used for the "mini-cell" that broadcasts and receives all signals from wireless network interface cards (NICs), also called wireless adapters. The AP connects to the rest of the wired network, via unshielded twisted-pair or optical-fiber cable.
By far, the most popular of the three standards is 802.11b, which operates in the 2.4-GHz band (also known as the ISM or Industrial, Scientific, and Medical band). It offers a maximum data-link rate of 11 Mbits/sec. 802.11b offers three non-overlapping channels, which means that any implementation that requires four or more channels will be repeating at least one carrier signal. 802.11b operates on a direct-sequence spread-spectrum (DSSS) encoding scheme. DSSS was chosen for its ability to operate efficiently in a noisy environment—a desirable trait since the 2.4-GHz band also includes microwave ovens, cordless phones, garage-door openers, and other RF devices.
The 802.11a standard operates in the 5-GHz band, which currently has minimal interference compared to the 2.4-GHz band. The specified data rates are maximum two-way equivalent values, and can range from 54 Mbits/sec to 9 Mbits/sec, and potentially to 6 Mbits/sec depending on distance and noise. 802.11a commercial systems offer eight non-overlapping channels, which allow for an eight-cell implementation before repeating of carrier signals.
The encoding scheme of 802.11a is orthogonal frequency division multiplexing (OFDM). It uses multiple sub-carrier frequencies and combines streams of data to provide more-efficient bandwidth use and more resiliencies to RF interference and multipath fading, when compared to DSSS. Systems based on 802.11a technology have started to gain momentum because of their capacity advantage over 802.11b. The cost of 802.11a equipment is roughly 2 to 3 times that of 802.11b equipment. The differences in encoding schemes and frequency bands between the two technologies prevent interoperability between 802.11a and 802.11b systems.
As with all radio transmissions, the higher the frequency employed, the shorter the range for whatever amount of power is consumed. Thus, frequencies in the 5-GHz band will be attenuated more quickly than frequencies in the 2.4-GHz band. Consequently, the radius of operation for 802.11a is smaller than that of 802.11b.
The 802.11g standard is expected to be ratified this year. It is designed for a maximum of 54-Mbit/sec data rate in the 2.4-GHz band, and employs OFDM and DSSS encoding; therefore, it is backward-compatible with 802.11b. Like 802.11b, 802.11g operates in three channels because there has been no expansion of the frequency band (2.4 GHz). Pre-standard products are now shipping with this technology.
Without question, WLANs provide significant mobility advantages. According to Fortune magazine's November 25, 2002 issue, about 90% of the universities in the United States have installed WLANs. Obviously, universities represent the extreme in mobile "workforce" prevalence, and the advantages of wireless connectivity to the network represent a huge benefit to student utility. Why, then, does nearly every university with a WLAN continue to use cabled networks in dorm rooms, libraries, and computer rooms?
The answer has mainly to do with throughput limitations. Each of the three variants of 802.11 (a, b, and g) operates in half-duplex. This transmission scheme, coupled with payload overhead requirements, effectively reduces the maximum usable data throughput to less than 50% of the standard's stated data-link rate. In practice, the maximum data throughput of 802.11b is about 5 Mbits/sec; and that data rate is shared among all users.
In August 2001, Atheros Communications conducted a controlled test of 802.11b systems and found they provided an average throughput of 5.1 Mbits/sec when the user was 65 feet from the AP. In the same range, 802.11a systems provided an average of 22.6 Mbits/sec. Realize, however, that in practice, if 20 users are connected to that AP, then those users are sharing its throughput capacity. Each user should expect, on average, to get an equal share of the throughput rate. In the case of an 802.11b system with 20 users, that equates to roughly 250 Kbits/sec—1/400th the throughput of a switched 100Base-TX system.
Because none of the 802.11 variants offers guaranteed bandwidth for individual users, any single user can create enough congestion on the channel to temporarily, severely limit the throughput of other users. How temporary depends on the application or file being transferred. A client downloading a 10-MB file could be contending for every available packet for several minutes, which could lead to blocking and a dramatic throughput slowdown for all clients served in the area. In an effort to avoid situations like this, random back-off timing intervals are built into the AP, with the intent of providing equal access to other clients, albeit at slower rates.
Interference is the other major factor influencing WLAN throughput. The throughput numbers cited above assume no interference. In practice, interference comes from two basic sources: co-channel interferers and foreign interferers. Co-channel interference (CCI) occurs when a different 802.11x AP, operating on the identical channel, broadcasts into the area served by the home AP. 802.11x systems manage CCI by slowing down the data link rate to a point where the interference stops causing bit-error-rates. Of course, any bit errors still incurred due to CCI result in retransmission, which also effectively lowers system throughput. In Atheros' August 2001 test, an eight-cell 802.11b system—considered representative of small and medium-sized enterprises—with no foreign interference exhibited a 50% decrease in throughput, when compared to a system with no CCI.
An 802.11a system, which allows for eight unique channels, is not impacted by CCI until a system is deployed with more than eight cells. While this is a clear advantage of 802.11a, that advantage is countered by the fact that the cell radius of 802.11a APs will generally be smaller than those of 802.11b or 802.11g systems.
Foreign interference can come from a myriad devices operating in the ISM spectrum. Microwave ovens, cordless phones, Bluetooth and HomeRF devices, and other machines generate RF noises to 802.11b and 802.11g systems. While CCI is a spread-spectrum form of interference and, therefore, less disruptive to 802.11 receivers, some foreign interferers are not spread-spectrum and, consequently, may generate narrowband RF spikes that can cause greater interference problems for WLANs. Of course, a company has the ability to control the types of interferers inside its own walls. It may not, however, have the ability to stop foreign interference from a next-door tenant or from immediately outside the building.
The relatively clean 5-GHz band gives 802.11a the advantage over foreign interferers, but this may also be short-lived. Cordless phones operating in the 5-GHz band are being introduced.
Security in the spotlight
Literally hundreds of articles and papers have been written lately about war-drivers, wireless hackers, and the efforts to block them. By now, most people with an interest in WLANs know that most of the publicized breaks in WLAN security have occurred because the security feature of 802.11 was turned off. Simple enough to solve, right? Not exactly. The security mechanism packaged with all 802.11 systems, called wired-equivalent privacy (WEP), is acknowledged to be fairly easy to crack. Some experts can steal the key in a few seconds, and software that can be used to break WEP is available for free download from the Internet. Because of the great concern over this security shortcoming, the IEEE is working on a new security standard, called 802.11i. This standard is expected to come out late this year. On the downside, it could obsolete some current WLAN equipment. In the meantime, proprietary authentication and encryption algorithms are available. Generally, they require wireless NICs and APs to come from the same vendor.
Currently, the acknowledged best method for creating a secure WLAN is to install a virtual private network (VPN). A VPN will provide a secure connection from the network to all wireless clients, and will work with any standard-compliant 802.11 systems. A VPN represents additional cost to a network administrator.
To calculate the actual costs of installing and deploying a WLAN, you'll need to go a little bit beneath the surface. The costs of APs and NICs are fairly self-evident; APs for 802.11a are currently under $400 and NICs are approximately $170. 802.11b APs are available for about $200 each and NICs are about $60.
Beyond those costs, determining how many APs you will need to optimize coverage will, in all likelihood, require a site survey—and its associated costs. Additionally, WLAN installation includes expenses for mounting APs, running cabling to each AP from a switch or hub, and installing new power outlets to support the APs. On average, figure $100 for the cost of cabling to each AP, and another $150 per AP for installing hardware for the radio and antennas. You may also need to include costs for installing wireless NICs in user devices and getting the application running.
It may be possible to save the cost of adding electrical outlets by using Power over Ethernet (PoE) technology. IEEE's 802.3 group is studying the subject. A PoE solution only requires a technician to run one twisted-pair cable to the AP to supply both power and data. A wireless AP can operate solely from the power it receives through the data cable. Prices can range from around $90 for a single-injection unit to between $400 and $500 for multiple-outlet power supplies.
Taking all of these costs together, and assuming 12 users per AP, the average cost per user of installing a WLAN is, conservatively, $115 for an 802.11b system and $238 for an 802.11a system. Other, optional technology—such as VPN technology and anti-intrusion hardware and software—can add to that cost. By comparison, the average cost per user for installing a Category 5e cable drop is about $100, including labor and materials. While the costs of wireless APs and NICs will most certainly continue to decline faster than that of wired networks, it likely will be a few years before per-user costs represent a significant savings for WLANs over wired networks.
This discussion has detailed initial costs of WLAN implementation. One of the main selling points of the WLAN is future cost savings that can be realized with every move/add/change (MAC). Clearly, it is more cost-advantageous to add one more person to an existing WLAN (for just the cost of a wireless NIC) than it would be to add a new Category 5e drop. On the flipside, adding voice application to 802.11 systems is a proprietary and expensive proposition. So, if the plans are for a new user to have a wired voice connection, then a cable-pull is inevitable. In that event, the incremental cost of adding a Category 5e connection to the wired voice connection is small.
A few companies are offering adjunct systems that provide voice application support over 802.11 technology. These systems utilize proprietary software to give priority to voice packets over non-voice packets (typically referred to as QoS—quality of service). Lack of QoS has been a reason that voice service has yet to take hold commercially. The proprietary solutions available today solve the voice packet prioritization problem, but require the addition of fairly expensive VoIP servers, proprietary gateways, and proprietary phone sets. The costs of these systems will certainly come down once QoS is standardized for 802.11, which is targeted for ratification later this year.
The final cost consideration is total cost of ownership over a system's lifetime. In a nutshell, the rapidity with which WLAN technology is developing results in the real possibility of current-generation equipment's obsolescence sooner rather than later.
With the advent of 802.11x, the industry has a set of wireless standards that provides capacity and ease-of-use at a reasonable cost per user. Many organizations can now justify the cost of 802.11 systems based on productivity improvements (as in the case of large engineering companies) or the utility that a WLAN promises (as many colleges and universities have done).
The two primary drawbacks of 802.11 systems are throughput limitations and security gaps. These two issues can be quite significant for private and public companies, although they probably are not critical to most WLAN home users. With the expectation that more capabilities soon will be built into 802.11 equipment, and the lack of voice-application support (within the standard), the potential obsolescence of currently installed APs and NICs gives some pause as well.
For much of corporate America, for the next year or two at least, WLANs will be most practical, and most easily justified, as an overlay to the wired network for users who need mobility to improve productivity.
Mark Farrant is director of new business technology and Tim Waldner is senior vice president of marketing for SuperiorEssex (www.superioressex.com).