Beware jumping the gun with 802.11n

Joanie Wexler, Network World

Throughput promise of 802.11n lies in 5GHz band
By Joanie Wexler

Earlier this year, the IEEE 802.11 Working Group approved Draft 2.0 of the 802.11n specification. This means that the spec is not likely to change substantially from now until the standard is ratified, probably in the October 2008 time frame, so compliant products should get under construction and out the door, fairly soon.

802.11n represents the next generation of Wi-Fi networking that will deliver significantly higher throughput – up to six times in the near term – through the use of multiple-input multiple-output (MIMO) spatial multiplexing technology and signal processing. How much throughput you get depends largely on how many transmitting and receiving antennas are on each end of the communications link. Early systems using two spatial streams and channel bonding techniques, which I’ll discuss in a minute, will likely get a theoretical maximum throughput of 270 Mbit/s to 300 Mbit/s.

You’ve likely heard that 802.11n is required to be backward compatible with existing 802.11a/b/g networks. This means that it will run in both the 2.4GHz band of 802.11b/g and the 5GHz band of 802.11a. The real promise of 802.11n, though, is in the 5GHz band.

5GHz has lots of non-overlapping channels (12 in the U.S.; up to 23 worldwide). Because of this flexibility, channel bonding will be most practical in the 5GHz band. Channel bonding is part of the 802.11n standard, and it allows you to bond two 20MHz channels together for double the aggregate throughput – sort of like the concept of N-by-T1 inverse multiplexing in the WAN discipline.

In the 2.4GHz band, you only have three non-overlapping channels. Generally, you overlap cells of coverage using the three different channels. Each non-overlapping channel in a (Chinese) checkerboard-style layout is far enough from the same channel in the next cell so as not to cause interference. When you bond channels in the 2.4GHz band, you reduce the number of non-overlapping channels from three to two. This means you have to repeat use of the same channel a third more often. This significantly limits your cell design flexibility, which could create interference. So channel bonding won't be used much in the 2.4GHz band, which inhibits 802.11n's potential for speed in that band (except in single AP set-ups such as home networks).

A few years ago, 5GHz spectrum rules were so inconsistent all over the world that it would have been difficult to develop a global strategy for 802.11a (the Wi-Fi standard in that band) etworking, even if client devices and handsets had been widely available. Ditto for 802.11h, the derivative of 802.11a that avoids interference with military radar, especially used in Europe.

At this juncture, however, global harmonisation is fairly complete, with all but a few Middle Eastern countries and Russia allowing 5GHz networking in at least one of the four 5GHz networking spectrum bands:

  • 5.15 - 5.25GHz (four channels)
  • 5.25 - 5.35GHz (four channels)
  • 5.470 – 5.725GHz (11 channels)
  • 5.725 – 5.825/5.850GHz (four or five channels)

Having some harmony among the rules is extremely helpful to worldwide equipment rollouts and will be particularly important as 802.11n enters the picture.

As you create your global Wi-Fi plan to include 802.11n, however, keep in mind that some countries have their own rules about operation. They might require you to have a permit, for example. Some countries prohibit the use of 5GHz networking in hospitals. Others have restrictions about running it outside, which certainly has implications for muni/metro Wi-Fi networking.

If you are planning to run 5GHz Wi-Fi globally, be sure the chipsets in the wireless products you buy can accommodate all the channels in all the countries where you will be deploying. From there, most vendors’ product guides will inform you of the idiosyncratic rules that vary from country to country.

What about equipment?

Some enterprise-class products complying with the current draft of the emerging IEEE 802.11n wireless LAN standard are poised to hit the market. Meru Networks has announced, for example, that it intends to ship access points and controllers this summer that comply with Draft 802.11n, and Colubris has announced access points for the autumn - which work without an upgrade to the switches.

Cisco meanwhile is keeping its cards close to its chest. It has a draft N access point on the list that will be certified by the Wi-Fi Alliance, but refuses to say (even under embargo) whether it will actually launch it at Interop as one might reasonably expect.

Why do it?

If you are not experiencing any bottlenecks with your current Wi-Fi network, is there any reason to believe that your wireless traffic loads will suddenly jump to require the 100 Mbit/s to 300 Mbit/s speeds supported by these new networks? Many enterprises haven’t yet accumulated the volume of Wi-Fi application traffic that would require the extra speeds of 802.11n and thus merit the added investment.

If you don’t need 802.11n’s extra bandwidth today, you might want to keep 802.11n in mind but begin replacing any 10/100 Mbit/s LAN access switches supporting wireless access points with gigabit-speed Ethernet devices as the switches become due for upgrades or replacement anyway. As the final 802.11n standard isn’t expected until fall 2008, by the time you install 11n products, they might already be compliant with the final standard.

On the other hand, if you think you might indeed need the 100 Mbit/s+ bandwidth now, do a cost-benefits analysis. Depending on number of users, their degree of mobility, your mix of Wi-Fi supported applications and your productivity expectations, it might be worth biting the bullet, making the investment now, and getting an extra 12 to 18 months of benefit from the nascent technology. If you can demonstrate that the return of deploying now will be higher than what you must spend for initially high-priced products, which you may need to software-upgrade later, that’s where the rubber meets the road.