Should Wi-Fi stick to the ceiling?
Where do you want your access points...
By Peter Judge, Techworld | Published: 19:00, 29 September 2004
Earlier this month, Aruba put the cat among the pigeons by suggesting that Wi-Fi installations could save money and work better if we all forgot about putting access points in the ceiling, and put them at floor level. But the debate might not end there - more radical approaches could put all the access points in the wiring closet.
A big change - or a big mistake?
The launch of Aruba's Grid Point architecture, even went so far as to include - though a deal with structured cabling provider Ortronics - an access point integrated into an Ethernet wall jack.
The issue of where to put access points is of interest to more than the interior decoration team. The industry has built up a set of reasons for putting access points in the ceiling - to do with coverage and interference. Up in the ceiling APs have a clear line of sight to most of the office floor, people and filing drawers won't randomly block them, and they can reach further.
All that no longer applies, we are told. Access points can now be had so cheaply that range is no longer important. Plug a load of them in at under-the-desk level, and they'll cover the space you need. Sure they won't reach so far, but you have more. And short range is actually a benefit, because Wi-Fi has a greater throughput at shorter range.
Who is saying this is just as important as the message itself. Aruba is one of the most visible start-ups in the enterprise wireless switch arena. It has invested a fair bit of money in selling ceiling-installed WLANs. While the company promises it will continue to sell and support its previous product, that product has to take a knock. The justification for the wall-mount model is essentially that it will eventually mnake ceiling-mounting obsolete.
Aruba can find some influential support for its change: analyst Craig Mathias of the Farpoint Group is in favour of dense deployments: "There is, they say, an inverse relationship between distance and throughput in wireless systems," he says, in an article on MIMO and the fast 802..11n stndard. "The farther you go, the slower you go until no connection is possible at all because of fading. Coupled with the use of spectrum above 5 GHz, where propagation is more problematic than at 2.4 GHz, this means we'll see denser deployments of infrastructure (access points) in the future."
No boy scout would do that!
But rivals are more sceptical.
"Putting an access point at the floor, or desktop level where it is blocked by steel filing cabinets, cubicle frames, and desk materials is in the What not to do when deploying a radio network chapter of the RF Boy Scout Handbook," says Alan Cohen, director of marketing at Airespace.
To him, Aruba's strategy smacks of pico-cells, a good strategy for Bluetooth, which only need to talk to devices in one cubicle. Deploying Wi-Fi like Bluetooth is going to result in interference between the two of them, he says. (what are the worst causes of interference - read here). "Given the proliferation of Bluetooth devices [eg handsets from partners like Alcatel] an AP at the bottom of the cube is going to be more susceptible to Bluetooth interference as it is placed right next to the phone/PDA/cell phone," says Cohen.
He also thinks that the grid idea will spoil the advanced features such as location and security monitoring that are currently popular. APs at ground level will have their view blocked by furniture he says, so users will still have to deploy plenty of APs in the ceiling. And the performance of the whole thing will be adversely - and unpredictably - affected by clutter in cubicles
Cohen reckons it's a pitch to sell more access points and keep the price of systems up - though Aruba's David Callisch counters that it keeps overall costs down by cutting up to $1000 cost of installation and maintenance per access point.
Wi-Fi based on Pico-cells will only work, says Cohen, if the vendor makes changes to the client to takes advantage of it: "From experience, having a network that provides this feature is useless unless the client is capable of taking advantage of it - in fact, it can even be harmful if the clients do not cooperate," he says.
This may be a teaser, of course, because elsewhere, senior marketing manager, Jeff Aaron says Airespace is considering denser deployments (as reported in Unstrung) - but they are definitely staying on the ceiling.
Just stick the APs in a cupboard?
All of which sounds like the kind of argument which will run and run, and it also begs the question of whether there might not be another solution.
One radical suggestion is not to put the access points in the ceiling or on the wall - put them all in the wiring closet. The apparently daft idea comes out of the niche field of signal distribution. Where mobile coverage is poor, some companies are buying systems that improve cellular coverage inside the building, with a network of signal repeaters, coaxial cables and passive aerials.
These system are particularly popular in financial institutions, and places where people rely on mobile email, says Jeff Kunst, vice president of marketing at MobileAccess: "We're going gangbusters in the financial industry," he says. "It's all driven by BlackBerry."
It has recently dawned on MobileAccess that these in-building wireless solutions could could equally carry Wi-Fi signals also. The company has begun selling the idea of distributing Wi-Fi over the same system of passive cables. The access points can then all sit in the wiring closet, on the same shelf, connected over the passive network to aerials throughout the building.
The first customers for a Wi-Fi network distributed over fibre are customers already using the system for cellular signals, such as Lehman Brothers. These customers can distribute Wi-Fi over existing distribution networks in a matter of swiftly, says Kunst: "One company had a cellphone system, and put Wi-Fi in in hours. They used the same cables and had a ubiquitous Wi-Fi network in a few hours.
It's more of a challenge to sell to new customers, but Kunst says the idea is catching on. "We're seeing Wi-Fi driving our business, not seeing it as an add on afterwards.
"Wi-Fi was not designed for large instituations," he says. "The challenges have mostly been solved - but they could be addressed better by our architecture."
Most Wi-Fi networks start out with low bandwidth, deployed for coverage, not capacity, and that stores up trouble, he says. "Users end up with lots of little active components on the ceiling. They have a good mean time between failure, but the failure of an access point is very difficult to fix."
Dense deployment just makes it worse!
In his view, dense deployment just exacerbates the problem, putting even more access points out there. Even if they can be reached without ladders, they have to be found and handled, and more components is likely to mean more changes.
Putting the APs in one place in a closet makes it easy to maintain them - and it is even possible to switch APs between antennas when required, making the system more flexible. One spare AP is enough to cover the network, says Kunst, making resiliency cheaper to provide.
The capital expenditure can be higher, as the network needs extra components and coaxial cable, but Kunst reckons that using fewer access points will reduce the cost. The system really scores on operational expenditure, he says, because it is much more maintanable.
He sees it as a repetition of the evolution of Ethernet, which ended up with all the devices in the wiring closet: "Passive devices that don't need to be managed, and have a MTBF or 50 years," he says, "are the only things you should deploy lots of."Where do you want to stick your APs? Join our Forum and tell us what you think of the options.