Air Time: 11a or 11g: Which One Is Best for You?

Should I deploy 802.11a on my wireless LAN?

That's a question that frustrates me immensely. But it's also onethat I'm happy to field, as I have on many instances over the pastcouple of years.

If you are referring to a home Wi-Fi network, the answer is almostalways no. The only caveat relates to users in multiple-dwelling units,where finding an available 2.4 GHz channel that won't interfere withyour network is challenging. On most home networks, the broadbandInternet connection is usually the source of performance bottlenecksso even 802.11b is typically adequate. And because 11a operates at5 GHz, its RF absorption characteristics are high, which means it won't make it through very many walls before degrading into RF mush. In the home environment, where user density is usually quite low, maximizingcoverage is the primary design objective. In most cases, using an802.11g wireless router in conjunction with a modern wireless-equippedcomputer will be adequate. If your home is very large, you may want toconsider a high-gain antenna on your access point (AP) or MIMO-basedproduct offering to provide better coverage.

If you are referring to an enterprise network, or any network where youmight expect high user density, the answer is a resounding yes. That'sbecause high density translates into high contention and high contentiontranslates into reduced performance. Although it's true that most enterprisevendors allow you to co-locate multiple 11g APs on non-overlapping channelsand load-balance between them, such a design greatly complicates thechannel planning model and makes it difficult to provide full coveragethroughout a facility without introducing co-channel interference, whichoccurs when multiple APs and clients operating on the same channel overlapcoverage areas. Co-channel interference is a sinister enemy, often lurkingalmost invisibly in the background but then raising its ugly head as moreusers or time-sensitive applications are added to the network. Today,many network managers aren't even aware of its existence because usersare currently accustomed to glitches on wireless networks. But they won'tbe in the future.

One common approach to solving this problem is to use a micro or pico-cellarchitecture by reducing the power output of the APs, thereby shrinking theAPs' coverage radius. Although such an approach can sometimes yieldfavorable results, it doesn't help much if your APs are whispering and yourclients are yelling at full volume. Even though the APs themselves may notinterfere with each other, clients in adjacent cells may interfere with APswith which they are not associated. OK, so why not turn down the power onthe clients? That's not an irrational approach, but there are two problems.First, there is no standards-based mechanism that allows an AP to instructa client to turn down its radio power level (Cisco offers this capability withCCX, but it is only supported on Cisco infrastructure). Beyond that practicallimitation, turning down the client radio power level also decreases the RF linkbudget, which may mean that clients connect at a lower data rate. If you tryto increase the power level to provide connections at, say 12 Mbps, that willincrease the RF propagation range, which in turn causes co-channel interferenceand contention with adjacent cells, even at relatively low signal levels. There'sjust no easy solution to this problem.

The biggest mistake network managers make lies in assuming that theyneed to make a choice between 11a and 11g. That's just not the case. Yes,it's true that the cost of a dual-band, dual-radio enterprise-class AP can be20 percent higher than an 11g model. But with the cost of APs falling, that'snot much of a price premium, especially when you consider that installationcosts usually far exceed equipment costs. Some argue that because the cellradius of 11a is smaller, they will need to install additional APs. That's true,but there's no reason you have to provide 11a coverage everywhere in yourfacility. Since 11a-only NICs are no longer sold, users can always fall back to11g if they are outside of 11a coverage. Even if only 20 percent of your usersconnect via 11a that translates into improved performance for every singleuser because it reduces contention in both bands.

Another common mistake lies in looking at the installed base of laptop usersand concluding that because only a fraction have dual-band radios it won't payto add support for 11a. If that's the case in your organization, then someonein purchasing is asleep at the wheel. With a price premium of $10 or so, almostevery organization should provision dual-band radios on their clients. Fortunately,notebook manufacturers are increasingly installing 11ag NICs on enterprise-classsystems by default, so the extra cost is invisible.

Once 11a client penetration increases, your life will be a lot easier. Channelplanning at 5 GHz is a whole lot easier because there are many more non-overlapping channels, which allows you to virtually eliminate the effects of co-channel interference. If no contention is tolerable, as might be the case if youare implementing wireless VoIP, you can also choose to implement WLANinfrastructure from vendors like Meru and Extricom, which utilize sophisticatedRF scheduling algorithms that not only eliminate contention but also makeinstallation planning much easier.

So remember, for enterprise Wi-Fi, think 11a AND 11g, not 11a OR 11g.1022