WORKSHOP

How Far Is It To 802.11 Wireless LANs?

by Dave Molta

With estimated revenues of between $150 million and $230 million during1995, the market for wireless LANs is a rather small one, especially whencompared to more established "wired" technologies like Ethernetand Token-Ring. Today's fragmented wireless market is a mix of proprietarytechnologies that offer fairly low performance at a high price. While wirelessimplementations have produced significant benefits in certain vertical markets,including retail, distribution and medical arenas, many network managersconsider the technology too exotic to be considered for widespread deployment.And the lack of industry standards that provide interoperability for multivendorwireless LANs certainly h asn't helped.

The wireless evangelists suggest that if everything falls into place, 1997may just be the year of the wireless LAN, with market projections of morethan $1 billion by the year 2000. One of the major reasons for their optimismis the imminent release of IEEE 802.11, which outlines the first standardfor truly open wireless LAN technology. More than five years in the making,with countless delays attributable to both technical debate and politicalinfighting, the IEEE is expected to have a standard in place soon, perhapsas early as November of this year. But is this technology that will putcabling technicians out of work or is it simply too little, too l ate? Wesuspect the answer is neither. The IEEE 802.11 standards will undoubtedlyinfuse the market with significant competition while making a wireless optionmore palatable to network managers who place a high value on interoperability.

The Basics The vision for wireless networking is a compelling one.Freed from the tethers of u nshielded twisted-pair cabling, we can wanderwith our notebooks or PDAs from office to conference room, floor to floor,and building to building, all without giving up high-speed connectivityback to our home server. How do we define "high-speed?" For wirelessLANs, you should expect a minimum of 1 Mbps. Doubling or perhaps quadruplingthat speed is not outside the realm of engineering possibilities, but increasedthroughput will almost always come at the expense of distance.

Distance from what, you may ask. In some cases, you may find wireless LANsoperating in a peer-to-peer mode, so the range limitation involves the distancebetween any two peer nodes. More often than not, however, you'll be usinga wireless NIC to connect to a network access point-a device that providesphysical connectivity between wireless and wired domains. In most cases,access points are MAC-layer bridges that connect a wireless LAN to an Ethernetnetwork, but it is likely that we will see other, more sophisticated accesspoint alternatives if 802.11 succeeds in expanding the market.

So how far away can you be from the access point while still maintainingconnectivity? That's a question that everyone new to wireless is sure toask. The short and simple answer is up to 1,000 feet. The more honest answeris that it depends on the quality of the "media," which, in thiscase, is the suitability of your physical environment for the transmissionof radio or infrared signals. In some "noisy" buildings, or facilitieswith reinforced concrete or metal walls, you may not get 200 feet. Thisis bad, right? Not necessarily. In fact, given the bandwidth limitationsof today's wireless LAN te chnologies, limiting the number of nodes withintransmitting or receiving range of an access point can actually result inbetter throughput, because fewer nodes will be contending for availablebandwidth. Viewed from this perspective, limited geographic coverage isa benefit of today's technology.

That probably sounds like spin doctor ing, and to a certain extent, it is.If you need to provide coverage in a large building like a hospital, you'regoing to need lots of access points. In today's market, where wireless NICsrun $400 or more and access points can cost up to several thousand dollarseach, freedom from wire comes at a fairly hefty price tag. But if 802.11is embraced in the market, inexpensive single-chip MAC implementations andcommodity radios may push the cost down dramatically. Just remember thatEthernet cards once sold for $895 each.

If it was only that simple. The IEEE 802.11 committee's laggard pace inarriving at a standard is only partially a result of political infightingamong committee members. The other half of the story involves the technicalcomplexity associated with implementing a wireless standard that deliverson at least a large portion of the promise. Issues like excessive powerconsumption on notebooks, hidden nodes that are "visible" to accesspoints but not to each other, and handoffs between access points all leadto a system that makes Ethernet look as simple as two tin cans and a string.

Let's Get Physical The IEEE 802.11 draft standard consists of a setof specifications that defines both the physical (PHY) and media accesscontrol (MAC) layers. The standard further defines two radio-based (RF)PHY standards and one optical-based infrared (IR) PHY standard. While bothRF PHYs are based on spread-spectrum radio technology operating in the 2,400MHz to 2,483.5 MHz industrial, scientific and medical (ISM) bands, the agreementto support two standards represents a political compromise of sorts.

Direct Sequence technology is the basis for AT&ampT's (now L ucent Technologies)popular WaveLAN system, which is used by a number of vendors, includingDigital, Persoft, Solectek, C-SPEC and others. The direct sequence spreadspectrum PHY uses a pseudorandom noise (PN) code that spreads data transmissionacross a fairly wide frequency domain. Direct sequence systems make veryefficient use of ava ilable bandwidth, and our testing suggests that generallythey offer higher throughput than systems based on frequency-hopping spreadspectrum, especially when the number of stations on the wireless "segment"is small. Transmission speeds of 1 Mbps and 2 Mbps are specified.

Although arguable, future 802.11 implementations are likely to be dominatedby systems based on frequency-hopping technology, currently used by Proxim,BreezeCom, Xircom and others. Unlike direct sequence, the frequency-hoppingPHY uses a pseudorandom hop sequence that spreads transmissions across bothfrequency and time domains. This strategy makes frequency-hopping systemsless susceptible to noise and to a phenomenon known as the "near-far"problem that results when multiple wireless LANs are sharing the same frequencyspectrum. Equally important, frequency-hopping radios are cheaper to buildand more efficient in power utilization-an extremely important factor fornotebook and PDA users. Like direct sequence, the freque ncy-hopping PHYprovides for transmission at 1 Mbps or 2 Mbps.

Outside the radio realm, the 802.11 draft PHY standard also supports pulseposition modulation over diffuse infrared for optical implementations. Thebase transmission standard calls for 1 Mbps operation with optional supportfor 2 Mbps. Diffuse infrared systems do not require line-of-sight for transmission,but the signals cannot pass through walls. This can be seen as both a benefitand a limitation. The limitations are obvious: Access is limited to a singleroom. The benefit is the flip-side of this limitation: Privacy and securityare much greater than radio systems, since eavesdropping is more difficult.

He y MAC, Where's That Node? Many network managers have a good understandingof Ethernet and Token-Ring MAC protocols. Compared to 802.11, those technologiesare simple. When you are dealing with wires, not only can you make reasonableassumptions about signal quality, you also have a clear understanding ofthe physical boundaries of your network. Throw all that out with wireless.Your media is invisible to all but the most sophisticated spectrum analysismeasurement tools, and the locations from which a device may be able tocommunicate to an access point are often unpredictable, highly dependenton interference from seemingly "innocent" devices like microwaveovens. Beyond these problems, there are other tricky issues the 802.11 committeeneeded to address.

The first was the need to devise a MAC specification that could accommodatethe coexistence of multiple logical networks in the same RF channel space.Closely related to this, mechanisms had to be developed to allow devicesin a large network to roam between access points without losing connectivityto the network. A particularly thorny issue involved a phenomenon knownas the "hidden node" problem. This occurs when two wireless devicesare able to transmit to and receive from a common access point but cannothear the transmissions of each other, as illustrated in the figure "TheHidden Node Problem."

The 802.11 MAC is based on a carrier-sense multiple access scheme similarto Ethernet, but instead of collision detection, 802.11 employs a collisionavoidance algorithm. Using a Distributed Coordination Function (DCF), theMAC (via the PHY) senses the wireless media for activity and transmits ifno activity is detected. Unicast packets require acknowledgments for eachframe transmitted. Because the hidden node problem makes it impossible toassure reliability of the standard carrier-sense algorithm, a request tosend/clear to send (RTS/CTS) control frame system can be used to reservebandwidth and thereby eliminate collisions, albeit at th e expense of additionalpacket overhead that can reduce effective throughput by up to 20 percent.

An optional Point Coordination Function (PCF) is also defined to supportcontention-free operation, useful for time-bound applications. Under thisscenario, one station acts as a point coordinator, controlling access tothe media through the use of a polling and prioritization scheme. The pointcoordinator controls access to the media for a duration of time known asthe contention-free period, at which point control of the media revertsto DCF. DCF and PCF defer to each other in a fair access manner.

In addition to the basic media access algorithm, the 802.11 MAC also includesprovisions for data security, roaming between access point service areasand power management. Security is handled via an optional Wired EquivalentPrivacy (WEP) system that provides privacy using a data encryption modelbased on the RSA RC4 PRNG security algorithm. Roaming is supported usinga system of association and re-association between wireless stations andaccess points, which communicate associations via a wired distribution systemthat interconnects access points. Finally, a management scheme allows idlebattery-powered stations to "sleep," and therefore preserve batterylife, using access points to buffer frames destined for the device. St ations"wake up" periodically and listen for beacons from access points,indicating they have buffered data waiting to transmit.

While 802.11 specifies multiple PHY options, all use the same MAC. Thisis critical to the success of the standard, since it results in the largestpossible market for single-chip implementations that are expected to reduceprices for wireless NICs and access points. AMD and Digital Ocean have announcedsingle-chip 802.11 MAC implementations designed to be incorporated intoPC Card, ISA and PCI NICs.

Now Here This By now, it is probably clear why the IEEE 802.11 committeehas taken so long to arrive at a standard for wireless networking. The endresult is a well-designed specification that accommodates the requirementsof most existing wireless vendors. A number of suggested enhancements, includingoptions for higher speeds, have been advanced by vendors, but it is unlikelythat these changes will be incorporated into the spec.

In March, the Wireless LAN Alliance (WLANA) was established by 13 suppliersof wireless LAN products and technologies, representing more than 95 percentof the existing wireless LAN industry revenue. The focus of this group willbe on raising awareness of wireless LAN solutions through an educationalprogram consisting of the sponsoring of educational seminars, developmentof position papers and general advocacy.

WLANA's major challenge will be to convince network administrators of thevalue of wireless LANs as a complement to traditional wired LAN technologies.In a market that is hungry for more and more bandwidth, successful advocacyof a standard that provides raw bandwidth of 2 Mbps and throughput of slightlymore than 1 Mbps will not be easy. The upside rests in the fact that manymainstream business applications are not bandwidth constrained, and carefuldesign of segmented wireless LANs can deliver adequate performance and mobilityunattainable in a wired network. If they are successful, volume shipmentswill likely push prices to a fraction of what they are today. With an openstandard that offers multivendor interoperability, 1997 just may be theyear of the wireless LAN.

Dave Molta is director of network systems at Syracuse University. Hecan be reached at dmolta@nwc.com. The author gratefully acknowledges theassistance of the IEEE 802.11 committee in providing materials used in thisarticle.

Updated May 31, 1996