I get asked this question a lot: With the advent of 802.11n, might it now be feasible to build the all-wireless enterprise?

Can we replace our wired networks with wireless and provision multiple services, primarily voice and data but perhaps video and more, on the wireless LAN? Are we now talking about not just the augmentation of the wired plant, but rather its replacement?

I recently observed a few experiments in an attempt to answer these questions, and the results were quite encouraging. But let me cut to the chase. No, we're not talking about the total replacement of all wire, just the wire at the edge of the LAN, where it meets the user. It's amazing how much wire is required for wireless, and with one small exception, that's not going to change. But for most users, yes, we're going all-wireless, and yes, we have the technology to make it happen today.

The small exception? Gigabit Ethernet. I think we can all agree that there is now little difference in performance between a wired LAN and a wireless LAN in most cases. We're still a few years away from gigabit wireless LANs, but for the bulk of typical office and enterprise applications, wireless LANs should be able to fill the bill quite nicely. That's assuming, of course, there is an infrastructure-side design and implementation appropriate to the mission of the enterprise under consideration.

The experiments I observed were part of a series of tests conducted by a medium-sized Silicon Valley company that's in the process of moving into a new headquarters. It was decided that the professional staff would go wireless to the greatest degree possible, meaning that notebook computers and telephones would be wireless and based on a common WLAN infrastructure. I've noted for some time that the RJ-45 and the RJ-11 jacks in cubicles and offices are mostly doomed, and this exercise really put this possibility to the test.

The building itself is two stories, each around 20,000 square feet, and mostly open space populated by cubicles and a few closed offices and conference rooms. It's the typical Silicon Valley setup. The objective was to provision service on the upper floor for a large number of data and voice users. In this case, the company tested a client configuration of 73 ThinkPad notebook computers, each equipped with a Cisco Aironet a/b/g PC Card adapter as well as 50 Ascom i75 handsets.

On the infrastructure side, a Meru Networks 3000 controller and 15 AP208 access points were used. The company wanted to know if there would be sufficient capacity to support both device classes simultaneously, particularly with respect to voice quality.

The experiments conducted were based on the popular IxChariot benchmark, and this tool was used to simulate both voice and data traffic. What was found? When testing throughput across the 73 notebooks, average performance of greater than 120 Mbit/s was found with all stations running simultaneously.

Simulating voice traffic on just these notebooks, there was an aggregate MOS (mean opinion score) of 4.29 (anything above 4 is considered excellent) and an R-value (another measure of voice quality) of 89.43, which is much better than most landlines. Adding the 50 Ascom handsets to this scenario yielded a MOS of 4.16 and an R-value of 86 - again, great voice quality. Voice quality sampled on some of the i75s was found to be excellent as well. Taking all of the Ascom handsets off-hook while running the performance benchmark did result in a loss of data throughput, but everyone agreed that the 64.748 Mbit/s performance was still adequate and that such a scenario (remember, this was 73 notebooks and 50 handsets all seeking service at the same moment) was highly unlikely. Overall, the company in question had no reservations in going all-wireless.

Note that, while the access points in question also included .11a radios, they weren't used. If they had been used, I'm sure the results would have been even better, since some traffic would have been placed on what amounts to another wire in the air. Also not tested was 802.11n, which is now becoming available. Ditto on the expected benefit here as well. But the company did get some hands-on experience with a large-scale Meru architecture. Meru has often noted that its products are designed for such installations, and I was impressed with the results.

It was interesting to note that the building where the testing occurred still had a lot of RJ-45s. It's hard, after all, to buy cubes without them. And what will the company end up doing with at least some of those RJ-45s? Well, apart from Gig-E, they make a dandy spot to plug in APs.

Craig J. Mathias is a principal at Farpoint Group, an advisory firm specialising in wireless networking and mobile computing. He can be reached at [email protected] This article appeared in Computerworld.