One from Column A, One from Column G

Coverage, capacity and cost – these are the three main variables when deciding which version of 802.11 wireless LAN technology is right for your network.

Coverage, capacity and cost – these are the three main variables when deciding which version of 802.11 wireless LAN technology is right for your network. While 802.11b currently dominates the Wi-Fi market, accounting for as much as 90% of the installed base, the industry is quickly moving toward two faster standards, 802.11a and 802.11g.

Both offer nearly five times the theoretical throughput of 802.11b (54M bit/sec vs. 11M bit/sec), while promising to deliver cleaner signals with fewer dropped packets, in part because of a more advanced modulation technology, Orthogonal Frequency Division Multiplexing (OFDM). That is where the similarities end. The major benefit of 802.11g is that it operates in the same 2.4-GHz frequency band as 802.11b and has been designed for backward-compatibility. The chief benefit of 802.11a is that it operates in the relatively uncluttered 5-GHz band and offers many more channels than 802.11b/g.

To further complicate the issue, vendors are developing multi-band products that include 802.11a, b and g technology. Paul DeBeasi, vice president of marketing at Legra Systems, says, “The industry is moving toward multi-band solutions. Customers are worried about the future of their networks, and they don’t want to be locked into any one standard. Especially in the enterprise, customers want multi-band flexibility, and chip vendors are beginning to deliver the silicon to meet that demand. In the near future, [network interface cards], chips and access points will all be shipping with multi-band capabilities.”

So does that mean companies don’t have to make the tough choices between 802.11a and 802.11g? Not so fast, DeBeasi says. “The question then becomes one of network design. Are you designing your network for an 802.11b/g infrastructure or an 802.11a infrastructure? Or, should you design for a multi-band deployment?”

In other words, even if your network has de facto multi-mode capability, this doesn’t mean it automatically supports both types of clients at once. Because a multi-band network is the most expensive option and most complicated to deploy and manage, most companies will settle on a single-band design. And that leads us back to our original question: What’s best, 802.11b/g or 802.11a?

The case for G: Goes with B, gets widest coverage, great cost advantage

If cost is your main decision point, then 802.11b/g is your best bet. While 802.11b has lost market share in recent quarters, it has mostly been to 802.11g. This mean that the lower prices that result from volume productions should be shifting to 802.11g’s favor.

According to Greg Collins, senior director at Dell’Oro Group, the migration from 802.11b to 802.11g accelerated in the third quarter of last year, especially in the small office/home office market. Dell’Oro found that while 60% of the access points shipped in the quarter were 802.11b, 802.11g grew from nothing a year ago to capture 39% of the market. Over time, Dell’Oro expects to see more dual-band units shipped, especially to large companies.

Operating in the lower-frequency band of 2.4 GHz, 802.11g provides a range of up to 300 feet, which is significantly better than 802.11a. The greater range translates into fewer access points for a given coverage area. Moreover, 802.11g’s backward-compatibility with 802.11b protects past investments in 802.11b-only laptops and handhelds.

The fact that 802.11g operates in the same 2.4-GHz band as 802.11b is at once its greatest strength and its most glaring weakness. Sharing the 2.4-GHz band makes backward-compatibility possible, but this is a crowded band. Everything from microwave ovens to cordless phones and Bluetooth devices can interfere with an 802.11b/g signal. Worse, because 802.11b/g provides only three non-overlapping channels, when interference is encountered on one channel, there isn’t much room to move.

Another key drawback of 802.11g is that supporting legacy 802.11b clients comes with a steep cost: performance degradation. To handle slower clients, 802.11g networks decrease their data rates when an 802.11b client is present, limiting access point throughput to the 11M bit/sec of 802.11b.

There is also some concern that 802.11g won’t be as backward-compatible as promised. The 802.11g standard contains a protection mechanism that prevents 802.11g clients from receiving preferential treatment over slower 802.11b clients. The mechanism is bulky, adding overhead to packets, and while 802.11g uses the high-speed OFDM modulation scheme, it must also support the legacy Complementary Code Keying of 802.11b, which further degrades throughput.

To address this problem, several infrastructure vendors have begun offering the Turbo G mode in their 802.11g products. To do this, they use a technique called channel bonding, while also turning off the 802.11b protection mechanism. In essence, Turbo G leaves 802.11b clients out in the cold, while having the unintended effect of generating interference that can affect the performance of nearby access points.

Judging from these problems, you might assume that 802.11g doesn’t stand a chance against 802.11a in the long term. “That’s not true,” DeBeasi says. “If you emphasize coverage over performance and you place a premium on good legacy support, then 802.11g will serve you well.”

Insert G here

Certain environments are well suited to 802.11g. DeBeasi’s company has been working with a large hotel. Serving a diverse customer base is the hotel’s greatest need, and an 802.11b/g infrastructure will cost-effectively accept the broadest range of devices.

“Keep in mind,” DeBeasi says, “even though from an infrastructure standpoint you typically want to choose between 802.11a and/or 802.11b/g, clients don’t have that problem.” With multi-mode capabilities rapidly becoming the norm on the client side, there won’t be that many, if any, 802.11a-only devices entering most networks, and, thus, 802.11b/g lets you serve the broadest possible client base.

Other scenarios where 802.11g makes sense are in conference rooms or corporate lobbies. In both situations, it’s impossible to know which clients will need access. Low-bandwidth applications are also good candidates for 802.11b/g, such as retail point-of-sale or inventory tracking.

According to Jeff Abramowitz, senior director of marketing for Broadcom’s wireless LAN products, 802.11g is driving the Wi-Fi market. “Infrastructure is now being upgraded to 802.11g, and going forward the question becomes whether the capacity gains of 802.11a justify the additional cost.” Abramowitz argues that network professionals are familiar with the 2.4-GHz band, they are looking to lower-cost products whenever possible, and they are drawn to the backward-compatibility of 802.11g.

The case for A: All about performance

The case for 802.11a is pretty straightforward: performance. “Some network managers are beginning to realize that as they deploy more nodes across the enterprise, they will run into capacity constraints,” Abramowitz says. “However, with 802.11a-based products being more expensive than the 802.11b/g products, due to lower production volumes, enterprises must decide whether the performance gains justify the added cost.” Adding to the cost equation is 802.11a’s lesser range – only about 180 feet. The more limited coverage area means that more access points are needed to cover a given space, resulting in additional costs.

Based on OFDM modulation, 802.11a delivers a raw data rate of 54M bit/sec and operates in the relatively clean 5-GHz band.

Significantly more channels are available in 802.11a than 802.11b/g; as many as 12. But the FCC recently opened more spectrum in the 5-GHz band, so the number should jump to more than 20.

The FCC also “harmonized” the unlicensed spectrum in the U.S. with that used overseas. In other words, your 802.11a clients could soon have international interoperability.

“Today, the number of 802.11a units shipping is negligible,” Collins says. “That will change, though, as more large enterprises begin deploying wireless LANs. What will drive the shift towards 802.11a will be raw data capacity issues.” Today, enterprise wireless networks are so new that network architects aren’t yet encountering capacity issues. Thus, the driving deployment issue today is investment protection, and in that scenario 802.11g wins out.

“That will change as the enterprise market grows,” Collins says. “We think 2004 will be the year of WLAN in the enterprise. The enterprise market grew 24% in [the third quarter of] 2003, and that growth will continue over 2004. As the enterprise market grows, you’ll see more 802.11a devices, but most of these will actually be multi-mode units. The healthcare, educational, retail and warehousing sectors will continue to be strong, driving the overall enterprise market.”

DeBeasi has seen a demand for 802.11a from certain market segments. “We’re working with a large financial company that has a number of users who frequently download spreadsheets and large documents. In this case, capacity is key, and they are best served by 802.11a,” he says. “High-tech and healthcare companies are also good candidates for 802.11a. If you need to exchange massive files or databases over your network, you can only meet those needs with an 802.11a design.

The case for jumping on the multi-band wagon: Flexibility, future-proofing

If you don’t want to be locked into one standard, multi-band solutions can alleviate the problem. A multi-band network supports 802.11b/g simultaneously. Today, multi-band capabilities come at a premium. While this could quickly change as multi-band chips hit large production volumes, the reality today is that they could add as much as 20% or 30% to the cost of your initial deployment.

“At the moment, there is a price premium associated with multi-band products,” says Colin Macnab, vice president of marketing at Atheros Communications. “But if you look at a deployment in the enterprise, the issue of access-point price is trivial. The biggest cost for the enterprise comes if they ever have to touch those access points again.”

What’s more is the fact that spectrum is a finite resource. Macnab argues that even if you chose an 802.11a infrastructure, as your network gets populated 802.11b’s three channels in the 2.4-GHz band will be worth their weight in gold.

Even if most users migrate upstream to 802.11a, a company might still need the channels available in the 2.4-GHz band. In a network with centralized control and quality of service, clients with multi-band capabilities could be segregated by task or user role. For instance, a user needing only e-mail access could be pushed down to the slower 802.11b channels.

According to Scott Lindsay, vice president of marketing at Engim, smart access point software is emerging that has the intelligence to understand what the client base looks like. “Today, you have two choices with single-mode solutions: overdeploy or leave someone out,” he says. “Intelligent multi-band solutions eliminate this problem, adjusting to meet the needs of actual users.”

These solutions also protect the network from slow clients. “Imagine we’re in a conference room with a number of other people. We’re all close to the access points, so we’re all associating at the maximum data rate,” Lindsay says. “What happens, though, when some guy down the hall associates with that same access point? Due to distance, he’ll only associate at 1M or 2M bit/sec, and the aggregate capacity of that access point will degrade to the lowest common denominator. With intelligent multi-band functionality, the access point can move that slow user onto his own channel. Then, not only can you meet the needs of more clients, serving both 802.11a and 802.11b/g clients at once, but you protect your bandwidth as well.”

According to Engim and Atheros, many multi-mode chipsets can run in 802.11a mode or 802.11b/g mode, but not both at the same time. You initially can set up the access point to be either 802.11a or 802.11b/g, but to switch from one to another the access points must be reconfigured.

Atheros and Engim argue that true multi-band functionality comes from the ability to simultaneously handle all types of traffic at the same time.

Network design issues

Then there’s the matter of network design, which is complicated because 802.11a and 802.11b/g networks require different designs. Access points for 802.11a can be placed closer together than 802.11b/g access points without interfering. “For most customers, what a multi-band infrastructure gives you is the ability to reconfigure the access points via software whenever your needs change,” Legra’s DeBeasi says. “However, if you want to deploy a true multi-band network, accepting any and all clients, what you really need to do is start with an 802.11a design and then add an 802.11g overlay. A centralized device, such as a wireless LAN switch, can then provide the intelligence to load balance clients from channel to channel and access point to access point to achieve deterministic performance.”

This is a subtlety that often gets lost in the multi-mode discussion. For true multi-band capabilities, you must design two networks. Certain access points will overlap – the only truly multi-band access points in the network – while the rest of the network will essentially function in an either/or, single-band mode.

So, basically, you have three basic choices:

•  802.11b/g design, which provides broad device and network interface card coverage, while currently being the most cost-effective option.

•  If you want a high-performance WLAN, 802.11a offers the best possible network capacity.

•  The most-expensive option is an 802.11a/b/g multi-band design that supports all available clients and provides high performance. Capital costs are higher for a multi-band network, but it could bring down operational costs over the long haul by saving you the trouble of upgrading down the road.