Wireless switches - a costly waste of time?
VC money has poisoned the Wi-Fi industry, says Cirond chief
If you haven't met Nicholas Miller of Cirond before, brace yourself. There isn't a single assumption about the wireless LAN market that he won't challenge. He see himself as a cross between the boy who said the Emperor has no clothes, and a feisty David to the (not actually very huge) Goliaths of the wireless switch industry.
That is some act, but as the head of the company that brought us the acclaimed WINc client, Miller's opinions are worth hearing.
He spent an hour with us at the WLAN Event in London's Olympia, explaining just why the enterprise wireless LAN market is - in his view - completely ga-ga, and why a down-to-earth approach is going to win in the end.
Why, in short, a $3000 software tool from a privately owned company has a better chance than a whole bunch of hardware products starting at $30,000, from venture-funded start-ups.
Never mind the argument over thin and fat access points, the whole idea of a wireless LAN switch - a specialised piece of hardware - is a mistake, an over-priced distortion created by the wonky mechanisms of venture capital, says Miller. Ultimately, managing wireless networks is just a software problem, and can be done far more cheaply than switch vendors say.
Cirond started with the WINc client, which runs on PCs and PDAs, sniffing the air and presenting the user with a choice of the wireless LANs in the area, testing the quality of each with a ping. The client stores the Wi-Fi connections the user prefers and connects to them when required. "It's the most popular wireless connectivity tool," says Miller, "with 50,000 copies in use."
WINc Manager extends that, running on a server, managing connections through access points The first version came out last year but, as , the gloves are coming off, and Miller is starting to present it as a wireless switch-killer.
Switches are a figment of VC money
There is not enough market for over-specified wireless switches, says Miller. "The whole wireless management market in large enterprises is only $150 million," he says. "Once the big guys, CA, HP OpenView, BEA and CA have taken their share, the others are fighting for the leftovers."
"Trapeze and the others are driven by venture capitalists," he says. "Their business plans are designed by VCs." This means they have to work in multiples of $10 million, because that's the kind of money that VCs have to invest. "They have to make $20,000 products and target the top of the marketplace."
To justify their investment, the start-ups set up big. They employ sales staff, and target large companies. "Some of these companies have a VP of European sales, and they haven't even got a product," says Miler. With that kind of overhead, their products 'have' to be expensive. "You can't have a salesman get into his Buick for a product costing $4000 or $5000."
The start-ups backed themselves into hardware and, to make matters worse, tied everything to the "thin client" argument, he says.
"The thin AP is going to go away, because the math doesn't work," says Miller. It might have worked if the price of consumer APs hadn't crashed., he says: "$800 'enterprise' access points are ridiculous. Take the lid off and they are based on the same chipsets." The whole idea of thin APs made more sense when regular access points were expensive, he says. Now, access points cost as little as $40.
"What it is about is a hardware game," he says. The switch start-ups are locked into selling - and justifying - their pricey hardware, when most businesses can get by with a set of commodity access points.
Under cover of the thin access point argument, switch vendors have actually come up with a bunch of different ways to hook up their own access points to switches (see Beyond the thin/fat spat). Miller wants to junk all that. He says the actual job of managing those access points is a software task (and if you press any marketing manager at a wireless switch company, they will agree, more or less, saying that the company's real IP is in software).
If you've spoken to any CEOs before you'll know what's coming. You've guessed, it's the cue for the spiel on the new version of WINc manager.
The Model T Wireless LAN
WINc Manager version 2.0, from the demo, looks like a basic wireless LAN management system that does pretty much everything a centralised wireless LAN hardware system does - and more than some - at a fraction of the price.
It does load balancing and planning, and RF management. It can take a CAD drawing of the office floor and model how the signals will work when you place access points across it (much like the Ring Master software included with the Trapeze Mobility System (reviewed here).
While switches put a lot of effort into deciding which AP should link to a given client, the WINc approach is to use the WINc client software to make that decision. "You need a client on every PC," he admits, "But you have that already, with your XP or your wireless card. And the XP client is junk."
Get real on security
WINc Manager has a take on security too. While the WPA security spec fixes holes in the earlier WEP spec, older laptops don't support it. Cirond's answer is to provide a function it calls AutoKey. "We generate a key and distribute it; the key is rotated as often as you like," he says.
WINc Manager can use the access point signals to triangulate, locating any client device (or rogue access point) to within two metres. "Detecting rogues is easy," says Miller. "Actually finding where they are can be harder."
A neat addition to security is that the system can use its location feature to block "wardriving" or the use of Wi-Fi networks by people outside the building. The system can be given the co-ordinates of the building walls, and programmed to refuse connections from any device outside them, says Miller. "You needn't worry about signal leakage," he says. "You have a virtual shield."
More channels, too
And while you are about it, how about increasing the throughput at a stroke? Almost in passing, Miller takes a swift kick at one of Wi-Fi's most accepted bits of wisdom: the "three channels for Wi-Fi networks" rule.
He accepts that to have absolutely no overlap between the signals, Wi-Fi networks can only use the three channels 1, 6 and 11. However, choosing 1, 4, 8 and 11 gives you four channels, for a price (in terms of channel overlap) that is worth paying, he reckons. "There's 5 percent overlap between channels 1 and 4, and between 8 and 11," he says. "That makes almost no difference compared with the benefit of having an extra channel." (If you want to see this argument in detail, see Cirond's White Paper on the subject).
It's not just a matter of more channels meaning more potential throughput. Four channels is more significant than that. You need four channels to do Wi-Fi in three dimensions.
Three non-overlapping channels are plenty for a two-dimensional set-up, where circular cells cover a big floor in a warehouse, for instance. When you have more than one floor to unwire, there will be places where three signals will come through from the floor below, and you won't have another channel to use.
WINc manager can do this with standard access points, says Miller, arranging the channels they use to give four usable channels.
There are limits!
It all sounds too good to be true, and yes, there are weaknesses in the Cirond approach. For one thing, the system, so far, assumes that all access points will be on the same subnet, which limits the size it can extend to.
But who cares? "Right now, people don't need roaming between subnets," says Miller. "And WINc Manager will have it later this year, anyway." Almost all WLAN installations are small enough to go on one subnet, he says. Apart from the obvious vertical markets, of hospitals and schools, big large scale Wi-Fi deployment is a fantasy,
"Even large companies don't need large wireless networks," he says. "For them, wireless is an adjunct, and that makes it just like a small business wireless LAN."
"The biggest market in my opinion is the small and medium business," he says. All the Wi-Fi volume sales are in homes right now, where the wireless network is put in to avoid the need to run cables. The next market is small offices, which will put it in much the same way. "It is easier, cheaper and faster to install than wires. Small companies constantly move to new buildings, and the wiring is always unusable." All these companies need is a way to hook up more access points cheaply and securely.
But who is Cirond anyway?
Cirond Corporation, from the little financial information on the Web, is a fledgling. Although the company's mission, and its chief executive, have remained unchanged over the last year (see this article from Business 2.0, May 2003), there have been some behind-the scenes rejigging.
The shares of Cirond Networks were bought late in 2003 by Exmailit - a company which set itself up on the perverse premise of email-to-snailmail conversion - which then changed its name to Cirond Corporation. "The transaction was essentially a reverse merger," explains Miller. "Cirond essentially went public by our former parent company, Cirond Technologies selling Cirond Networks (which is where we hold all our IP) to ExMailit, which immediately following the merger changed its name to Cirond Corporation. The operating company remained the same, and all the management and employees stayed the same through the transaction."
The company's filings suggest an outfit whose financial performance has so far been at "noise" level, but which is planning to get at least some of the VC money that Miller despises.
The proof of this pudding - like all Wi-Fi products - will have to be in actual tests of the product. The company has a joint marketing agreement with Netgear, which may bring it to wider attention.