There's a kind of frenzy in enterprise Wi-Fi, as vendors compete to give their different spin on what 802.11n Fast Wi-Fi will do for us.
802.11n - due to become a full IEEE standard by the beginning of 2009, promises fast reliable connections to rival wired Ethernet. It's set to sweep the board in single-access point consumer equipment, and analysts Craig Mathias of Farpoint and Paul DeBeasi of Burton Group expect it to do the same in the enterprise. But sceptics are emerging that express doubt over what 802.11n can do for business.
One of the big promises of 802.11n is the wide open, interference-free spaces of the 5GHz band. There are 24 channels there, compared with the cramped confines of the 2.4GHz band where, as we all know, there are only three channels that don't overlap with each other.
This week, I've found that some people believe that 802.11n will suffer exactly the same limitation. It will, they say, have only three non-overlapping channels. I think they're wrong - but let's go into the reasons.
It's not a strict apples-to-apples comparison (we're not talking the same kind of channels), but it's important because the history of enterprise Wi-Fi is a process of trying to cover office buildings with a lot of access points, despite the limitations of having only three non-overlapping channels in the 2.4GHz band.
If two adjacent APs use the same channel, they will interfere with each other. To cover an office floor, you need three non-overlapping channels - and careful RF planning.
Trapeze, Aruba, Colubris, Cisco and others have made a living out of this process, but with the emergence of 5GHz 802.11n, Cisco's wireless CTO Pat Calhoun has said, more or less, that we can say goodbye to that, and companies should not use 2.4GHz for enterprise wireless, but move to 5GHz.
Unfortunately, according to Extricom's CEO, Gideon Rottem, that won't solve the channel interference issue, since 802.11n will only offer three non-overlapping channels.
Rottem gets this by a process of whittling away. Firstly, some of these channels overlap. Even more overlap if you are using the fast, wide 40MHz channels that 802.11n allows. And there are DFS rules - whereby Wi-Fi kit has to avoid frequencies used by radar equipment in the 5GHz band.
So Cisco's Aironet 1250 access point data sheet says "up to 24" non-overlapping channels, but there's a footnote that says the maximum is nine, if 40MHz channels are used, and warns this may vary by regulatory domain.
Rottem says the channels will be reduced further to three - and calls on analyst firm Gartner, to back him up. In July, Gartner issued Key Challenges Arise for 802.11n Deployments, a 12-page report which warned that putting 802.11n into enterprises may be trickier than expected - especially if there is an installed base.
There's lots of detail about 802.11n implementations in that report, which we won't go into, but one part Rottem refers to does indeed say that there may be only three 40MHz channels.
It's based on a conservative assumption, though. Back in July, it seemed that a big chunk of the 5GHz band, from 5.4GHz to 5.7GHz, might not be available, under FCC and other regulators' rules, because of new Dynamic Frequency Selection rules called DFS 2 that require WLANs to dodge away from frequencies in use by radar systems.
DFS has been a tricky issue, and earlier this year it https://www.techworld.com/features/index.cfm?featureID=3271"> threatened to hold up 802.11n in France.
In the US, FCC rule # 15.407(h)(2) requires that products operating in those bands must support DFS, says Michael Coci, director of technical marketing at Trapeze, a Wi-Fi vendor that's selling 802.11n products. The requirements have been mandatory in the US and Canada since July, and will be mandatory in May 2008 in Europe.
"From the regulatory perspective, those channels are certainly available," says Coci. "The other half of the equation is the vendor implementation." Pre-11n products, including the little-used 802.11a, used chipsets that weren't designed for DFS 2, so they have taken a simple way out - just not using those bands: "The vendor solution for newly shipped 11a APs is quite likely to be disabling those channels."
In July, Gartner simply assumed that this would continue to be the case, reducing the number of channels available. Coci says the situation has changed: "For 11n chipsets, the chipset vendors were well aware of the DFS 2 requirements and implemented them in silicon, thereby enabling use of those channels."
Gartner's Ken Dulaney agrees, pointing out in an email to Techworld that vendors including Cisco do already support DFS 2. He warns against Extricom and Meru who "are trying to sell their architectures as the only way to get full bandwidth with limited channels."
Opinions vary on how many channels you have in the DFS 2 enabled 5GHz band, with Gartner suggesting nine, and that ought to be enough for anyone. July's Gartner report said that the shortage of channels might "force a number of third-generation vendors to move to the fourth generation" - a term for the blanket architecture of Meru and Extricom.
In fact, "fourth generation" is widely used by Meru, and it's worth noting that a Gartner WLAN analyst, Rachna Ahlawat, joined Meru as vice president of strategic marketing in July.
Since them Dulaney has stepped back from suggesting there is any channel shortage that might drive users to blanket WLANs in the 5GHz band - though he does point out that blankets will be the only way to cover buildings with 40MHz channels in the narrower 2.4GHz band, since there's only one non-overlapping 40MHz channel there - but Gartner no longer recommends 2.4GHz for enterprise Wi-Fi.
He also points out that blankets should make it easier to adjust a network to the unpredictability introduced by 802.11n's MIMO, which uses reflections to create multiple channels: "You lay out the APs in a symmetrical pattern and adjust later," he says. "And they [Meru and Extricom] can adjust without having to worry about co-channel interference."
So in conclusion, it looks like there isn't a shortage of bandwidth in the 5GHz band - which probably means it's not a sin, as Rottem sugests, to use it for coverage.