Column It's always been a bit of a mystery to the average corporate PC user: why, if Wi-Fi runs at nearly 60 megabits per second, can you still plug several Wi-Fi access points into a 100 megabit Ethernet socket and drive it from your 8 megabit broadband ADSL router?
Shouldn't you need several 100 megabit lines - one for each access point? And why does it matter, anyway, since most websites run at about half a megabit?
So this week's announcement of a "pre-n" wireless aimed at the corporate market by Meru Networks may take a bit of explaining to the chief financial officer, or whoever is meant to sign the cheque to upgrade the firm's wireless networking - because, for the first time, that poor CFO is going to be faced with a request for upgrading the entire office Ethernet to GigE.
And that poor CFO is going to say that the sums suddenly don't add up. Meru will be followed by a cascade of other 11n copycats, probably starting with Cisco/Airespace this summer, as corporate suppliers switched from their previous pose of "802.11n can wait till the standard is ratified" to "Oh, crap, our rivals are shipping - get something out the door! FAST!" - but even then, an awful lot of people are doomed to disappointment.
And the main reason they're doomed to disappointment is the proliferation of Wi-Fi mobile phones. It's all very well saying how fast Wi-Fi can run. The question is how fast it will run in reality - once pulled off the enthusiastic pages of Aruba and Trapeze and Meru's websites, and stuck in offices with concrete walls and multiple floors and neighbours on either side, all with their own wireless. And, with an increasing number of things like the Nokia E61.
OK, get your calculator out, and watch me get lost in hundreds, tens, and units, because this is dead hard sums. First question: how fast does a Wi-Fi phone run?
The official answer is that Wi-Fi phones run at 11 megabits per second, and that's a lie for a start.
Wi-Fi phones use the oldest Wi-Fi standard still on the market: 802.11b and "b" has a theoretical maximum speed of 11 megabits. Or, if you actually plug it in and turn it on, roughly one megabit per second, if you're the only user, and right next to the access point. That's because the actual throughput in the standard is the radio data modulation speed divided by a complex figure - a figure which gets bigger if there is more than one access point in range, there is more than one client device attached to the access point, or the client device is further than 10 feet away from the point.
The thing about Wi-Fi today is that most computers don't use 11b. They have 802.11g radios in them, and most access point wirelesses are also 802.11g. Theoretically, this combination will run twice as fast, at 22 megabits, as 11b. That is also a lie: again, because in both cases that's the raw modulation rate, while in reality you're looking at everybody sharing a cargo-carrying rate of around 10 megabits. It's a lot quicker than 11g, but it really isn't 22 megabits.
That is, it would be a lot faster - until some idiot walks into the room with a b device. At which point, the AP has to slow down. It has to build in a delay protocol before sending a packet, and that delay was a trivial one in the first 11b spec, which ran at two megabits. But it's an awfully long delay if you're running at 22 megabits modulation.
Several Wi-Fi suppliers have found ways to dodge this problem, and under ideal conditions you can appear to show 11g devices running at full speed even with 11b wireless in the room.
But in a busy environment, the dodges quickly run out of air. I've played with some wireless systems which claim 90 megabits per second for their 11g radios. This isn't a comparative review, so I'll summarise: if the shop won't take them back, use the recycle bin. Even under ideal conditions they tend not to run at more than 40 megabits, they tend not to work with rival hardware, they tend to blot out wireless reception for everybody else in the neighbourhood, and they tend to simply ignore the presence of 11b devices anyway. Also, they tend to be unreviewable crap.
All this changes, more or less, with the 11n standard - using a new MIMO (multi input, multi output) wireless system; and when the IEEE finally ratifies 802.11n in 2009, it will include MIMO in the standard. That's the basic system now appearing in the open market, with the possibility of minor changes in the IEEE committees, all probably fixable with a software update.
And, for the first time this technology - coupled with channel bonding - means it is really, not just theoretically, possible to get more traffic through a Wi-Fi access point than through a 100 megabit Ethernet switch. It will really be the case that an 11n access point will handle more users at a higher rate than an Ethernet switch rated at 100 megabits.
Which explains why Meru is doing a big propaganda campaign to get corporates signed up to the new technology, because Meru believes it has the answer: the "virtual cell" - which means you can forget about the wired network and go all-wireless.
You can get all the information you need about virtual cells, on a technical level, from Meru Networks direct. Suffice it to say, here, that it makes planning your office network easier, because it means the net looks like a single access point on a single channel, with the same service ID - and so "roaming" from one to the next is far simpler, without massive work done planning the layout. Normal wireless networks have to avoid overlap; the virtual cell effectively eliminates that problem.
It's my own judgement that Meru has leapfrogged the competition here. That's not to say that rivals like Trapeze and Aruba can't compete, just that they have to polish up their PowerPoint presentations to cover this threat. But all of them are, still, in the same boat when it comes to the problem we started off with: Wi-Fi phones.
The thing about 11n is that it uses ferocious amounts of air spectrum. Earlier versions were positively anti-social, using reflected signals and multi-path channel bonding tricks to soak up every bit of available bandwidth and, in the process, blotting out any 11b and 11g devices in the area.
Meru believes it has a solution to that, but if you talk to its engineers it quickly emerges that they're assuming two things, which may be optimistic.
First, they're assuming that people will use the 5 GHz spectrum normally reserved for 11a. And second, they're assuming that 11b is obsolete and will go away any day now.
The 11a problem is a story for another day. 11n does, in theory, have the ability to select all Wi-Fi channels; the 2.4 GHz channels one through 13 as well as the wide range of 5 GHz channels. Normally, 11b and 11g are stuck in the 2.4 G spectrum; 11a is stuck in the 5 G spectrum.
But 11n can switch between them (although, not in a neat, seamless, virtual cellular way) and so if it finds that there simply isn't any available 11b or 11g spectrum it jumps up and uses the high frequency.
Back to our Wi-Fi phone! Right now, every Wi-Fi phone in existence uses 11b, not the faster 11g. And there's a simple reason for that: battery power. The 11g technology gets its extra speed from a modulation system called orthogonal frequency division modulation, OFDM - a technology that applies to WiMAX too. It works, but there's no free lunch. To get more bandwidth, you have to use more power.
Take a Nokia E61 Wi-Fi phone. It's the basic Truphone toy and I've been happily making VoIP calls over my office Wi-Fi system using it for the last three months without any ill effects. But there is a problem: you have to recharge the thing every night if you use the internet phone call ability. Restrict it to cellular, and the battery lasts most of a week; but even sitting in a Wi-Fi hotspot means you use battery power at four times the rate. Talk, and it obviously burns the battery up faster.
Nobody is going to predict how much battery an 11g Wi-Fi phone would burn, but one engineer described it as "at least double" and then said "and that's probably a conservative effort, assuming that someone actually develops a low-power 11g chip, which they haven't yet." Which probably explains why there aren't any. Imagine a Wi-Fi phone that flattens your mobile phone battery in half a day? Nobody would buy it.
Which means that if Wi-Fi mobiles increase - as everybody seems to believe they will - then 11b is not going to go away. And that means that in the 2.4 G field where 11b operates, throughput will remain low as long as there's an 11b device switched on. The whole network runs at 11b speeds while they're operating in most instances.
So ignore all the optimistic predictions about how fast 11n will go until you've actually tested it. Make sure it will run in both 2.4 and 5 GHz spectrum areas (not all 11n devices do). Make sure it will support your 11a devices if you have any. And test it while your Wi-Fi phone users are making internet calls.
If it can jump through those hoops, buy it. I'm betting a lot of 11n gear won't pass those tests for another two years. ®