Carrier WiFi

Open, Unwired Design

When he was announced as the new dean of the School of Natural Sciences and Mathematics (NS&M) at the University of Texas at Dallas, Dr. Myron Salamon said, "I was attracted to UTD by its drive to become a major public research university and, in particular, by NS&M's pivotal role in achieving that goal." He continued: "My aim is to make the school a model for 21st century academic research and education."

The centerpiece of that effort is the new Natural Science and Engineering Laboratory at UTD, scheduled to be finished late this year. Being built at a cost of $85 million, the new facility is funded by a joint private/public effort dubbed "Project Emmitt" -- a multi-billion-dollar economic-development agreement announced last year by Texas Instruments (TI), the Texas state government, and the University of Texas System. Project Emmitt includes the new $3 billion semiconductor plant being built by TI in Richardson, not far from UTD. And the new NS&E building, in turn, will be a showcase and testbed for the new state-of-the-art wireless network that, on its scheduled 2008 completion, will blanket the UTD campus with ubiquitous WiFi coverage, including VOIP capability.

The new, interdisciplinary lab will feature an open-plan interior design with common areas intended to foster interaction and collaboration among scientists and students in different fields. That design is based on the recognition that the lines dividing science, technology, and engineering are increasingly fluid.

"Engineering and some of the sciences are so connected these days," Bob Helms, the Dean of engineering and computer science, told the campus newspaper, the Mercury, earlier this week. "Where does one stop and the other begin? The boundaries are very mixed together."

'Screaming Out of the Gate'
Advancing the notion of unlimited interaction will be the wireless network permeating the new lab, says Mike Griego, the WLAN project manager for UTD. "The new lab is going to be the centerpiece [of the campus-wide network]," Griego says. "That building is very interesting because it's a result of the partnership between UTD, TI, which is just down the street, and the state of Texas. They really pushed for a state of the art network in there."

While many large universities are now pushing for campus-wide wireless coverage, both to facilitate learning and attract top-flight faculty and students, UT-Dallas was in the forefront of the wireless transformation as far back as the late 1990s. The university installed a network using early-generation Lucent Technologies Inc. (NYSE: LU) equipment that would eventually extend to cover some 80 percent of the campus.

"A few years ago we were in the forefront in the .edu space [in terms of wireless networking]," observes Griego. "We really came screaming out of the gate with this."

As the Lucent equipment aged, and Proxim Wireless Corp. took over that product line, it became apparent that retrofitting and upgrading the network made less sense than installing a brand new system. UTD looked at equipment from several vendors including Cisco Systems Inc. (Nasdaq: CSCO), which became prominent in the wireless LAN space through its acquisition of Airespace. Ultimately, the "virtual access point" technology from Meru Networks Inc. tipped the balance in favor of gear from the Sunnyvale, Calif.-based wireless networking provider.

"Since we were looking at VOIP as one of the primary applications on the wireless network, the way that Meru's virtual AP handled the traffic really caught our eye," Griego explains. "As far as I know none of the Cisco/Airespace products do anything like the virtual AP -- they don't do the single-channel, highly orchestrated access to the medium that the Meru system does."

Live Animals, Too
In the most recent fiscal year UTD spent just under $125,000 on the new network, Griego says, averaging around $600 per access point after educational discounts.

"As universities build out their communications infrastructure, most want pervasive network access," says Meru vice president of marketing Michael Tennefoss. "It's part of the basic business and connectivity required for the students and faculty. And when they look at ongoing costs, by going wireless and deploying it pervasively, there's greater operational savings and lower total cost of ownership the more broadly wireless is deployed."

In fact, UTD has estimated that it will achieve full return-on-investment within three years -- which would be a remarkable feat in academia, where wireless access is considered a "must-have" feature rather than a competitive differentiator.

What's more, the new NS&E facility will present some unique challenges to Griego and his team. "They're doing some interesting things in there -- it'll have a vivarium," (a facility where animals are housed for observation in environments resembling their natural habitats) he chuckles. "It'll definitely have some interesting RF challenges."

— Richard Martin, Senior Editor, Unstrung

wirelessfreak 12/5/2012 | 3:47:06 AM
re: Open, Unwired Design Last news from Japan (Nikkei Communications) is that the Osaka Gas deployment was forced to go multi-channel because of major issues with the virtual AP.

If Meru uses multi-channel do they still have the same benefits?
homeless_dad 12/5/2012 | 3:47:01 AM
re: Open, Unwired Design What kind of major issues?
wirelessfreak 12/5/2012 | 3:46:56 AM
re: Open, Unwired Design 1. Scalability and managability of virtual AP beyond a smallish number of APs
2. Roaming and redundancy between switches
3. Unable to avoid interference on the virtual AP channel
4. Total throughput limited by client intereference - if one client is transmitting at high output and other APs and clients can hear it then the total throughput of the virtual AP is severely limited

In the end the deployment required moving to multiple channels thus invalidating the major premise of Meru's architecture.

harrydbrownjr 12/5/2012 | 3:46:50 AM
re: Open, Unwired Design Are you refering to this article?


Latest News on Osaka Gas Co., Ltd.

Meru Unveils Call Load Balancing and Readies Radio Switch
Voice-over-wireless LAN (VoWLAN) developer Meru Networks Inc has unveiled a module for its operating system that enables a network to handle or refuse calls depending on resource availability, as well as call load balancing, and is now readying technology to blanket multiple channels of coverage to avoid network failures.
Published 1 August 2005 - News

MichaelRuiz 12/5/2012 | 3:46:34 AM
re: Open, Unwired Design I think someone really missed the point about the Meru Architecture.

Meru isn't saying that they've created bandwidth out of thin air. They are using the spectrum more efficiently, thus making reality closer to theory. (Theory being near 54Mbps on g, reality often being well less than half depending on lots of variables).

Virtual Cell technology coordinates clients and AP's in "earshot" of each other. Rogues and other noise can still interfere and that uses spectrum and uses bandwidth.

Imagine though a virtual cell on channel 1/g providing the best use of 1/g there is out there (results and tests prove this). When you have enough traffic/clients/density/etc that you are at the capacity of that channel, you add another virtual cell (second radio in ap, new ap, radio switch) on channel 6/g. And so on.

With Meru's technology you now have the ability to take real advantage of the full spectrum everywhere, not based on geography and channel overlap.

Mike Ruiz
Network and Systems Engineer
Hobart and William Smith Colleges
wirelessfreak 12/5/2012 | 3:46:28 AM
re: Open, Unwired Design The article about Osaka Gas was in Japanese and from Nikkei Communications, but I cannot seem to find an online version to reference.

The issues the article discussed are operational and scalability issues in addition to the RF "magic" that Meru attempts to perform with the virtual AP. The magic also evidently did not work as advertised and it seems, based on Michael Ruiz's comments, that Meru is moving away from the single channel, single BSSID approach to a multi-channel approach.

Thus, my original question - does the virtual AP actually work at this scale? and what features are available or not when moving to multi-channel and multi-switch (e.g. does zero handoff still work as advertised? how is roaming across switches? how is redundancy with multiple switches?)
lrmobile_rusty 12/5/2012 | 3:46:26 AM
re: Open, Unwired Design I come from a training background so scalability questions are out of my area of expertise. That said, the multi-channel approach still uses a single BSSID for each channel. The idea is that you put multiple radios (I think up to 4) in each AP so that you have 4 separate virtual APs overlaid on top of one another.
MichaelRuiz 12/5/2012 | 3:46:24 AM
re: Open, Unwired Design Find me a source on the "Issues" at Osaka and I will get the answers though.
MichaelRuiz 12/5/2012 | 3:46:24 AM
re: Open, Unwired Design Thanks Rusty... You're right in that the "Multi-Channel" approach is not a new approach for meru. It is multiple layers of their Single-Channel architecture. One BSSID Per Channel load balancing, etc.

The system is absolutely scalable, by design. I have over 200 Access points with 2 radios per. I have only a few locations where I've added a second cell for more bandwidth.

I know of several installs that are significantly larger.

Also the Radio Switch has a model capable of up to 12 radios all capable of simultaneously using a single patented antenna.
frnkblk 12/5/2012 | 3:43:41 AM
re: Open, Unwired Design It could be all targeted marketing, but when I polled .edu's for dense wireless deployments (more than 50 clients per AP), Meru was the name that kept popping up. According to all the interviews the Meru stuff 'just worked', replacing the products from traditional vendors which did not work as well. When I polled the other vendors for similarly dense deployments it was hard for them to come up with any that environments that daily matched or exceeded the Meru-supported dense wireless scenarios.

As Michael Ruiz pointed out, where Meru needs more bandwidth they recommend their RadioSwitch. I don't think it invalidates Meru's architecture (VirtualCell with controlled access to the medium), but adding radios will increase cost. It's not been widely deployed, so I reserve my judgement on that count.

I had heard the same thing about the Osaka Gas deployment, but I forget what Meru's response to that was. In any case, there are many deployments with Meru that can confirm that the product scales, roams well (I'm not sure about between APs on different switches myself, but I'm sure Ruiz can comment on that), and provides effective throughput in asymmetric power situations. It would be helpful to hear from an ex-Meru customer on why the solution wasn't effective for them.

Aruba is dealing with the density issue by recommending a 4-channel configuration (1, 4/5, 7/8, and 11). Adding just one more channel really helps with channel design, but there are still the adjacent channel issues to keep in mind. Other than that, reducing client and AP output power and using micro-cells are the only current methods to increase density, but it does increase the number of roaming events and complicate channel design in the 2.4 GHz range.

Sign In