University seeks Ethernet-like wireless behavior

* Finds consistent symmetrical performance important for staff use

Last week I had the pleasure of speaking with a network project leader at a Swiss university. Ha Nguyen of the Universite de Lausanne, near Geneva, Switzerland, displayed a practical attitude toward 802.11n deployment issues as he discussed the symmetrical nature of true business traffic and the real reason for the latest centralized vs. distributed architecture debate: cost.

Last week I had the pleasure of speaking with a network project leader at a Swiss university. Ha Nguyen of the Université de Lausanne, near Geneva, Switzerland, displayed a practical attitude toward 802.11n deployment issues as he discussed the symmetrical nature of true business traffic and the real reason for the latest centralized vs. distributed architecture debate: cost.

For about six years, the 12,000-student university has run a best-effort WLAN using Cisco legacy stand-alone access points (AP) in public areas such as cafeterias and libraries. Last year, amid the hubbub surrounding emerging 802.11n technology, the university began to investigate new alternatives for full coverage, in part because staff was asking for mobile connections. Nguyen found 802.11n to be a pretty costly proposition.

As a result, one of his criteria became a distributed architecture. “You can get good performance with any architecture [distributed or centralized],” he said. “But at what cost?” He wanted to avoid controllers that backhaul user traffic because he found that controllers that were powerful enough to handle gigabit wireless LAN traffic were cost-prohibitive.

Beyond that, his approach was pretty simple. He believes that the wireless “weak link” is the radio itself. “So we try to keep a short list of vendors planning to [innovate] in the radio,” he said. And he asks vendors to send him sample products so he can test them – a necessary, but costly, practice, he acknowledges, because Nguyen tends to take the paper claims of vendors with a grain of salt.

For example, to accommodate 3,000 university staff, he accounted for the nature of their business traffic flows, which tend to be far more balanced (about 40% uplink and 60% downlink) than that of students browsing the Internet (about 90% downlink).

“Client/server applications are pretty symmetrical,” he said. “If I have a wireless PC and I mount a disk on a server, I’m going to read and write – not read only. When sending e-mail, I’m attaching files to my uplink. When groups of engineers work together on a project, they exchange information.” So he needs the WLAN to be very Ethernet-like in its behavior.

But the vendors he evaluated with airtime fairness algorithms apply the mechanism in the downlink direction only, he said.

He has installed a 20-AP testbed of Ruckus Wireless APs and plans to install 500 to 1,000 this year “because we get more stability in performance” largely due to the antenna-based beamforming capabilities of this particular system.

Not coincidentally, Ruckus formally entered the high-end enterprise market earlier this week with the announcement of its $999 dual-band Draft 11n AP, which the company says was purpose-built for 802.11n spatial multiplexing. Its antenna-based beamforming capabilities choose best paths to clients based on thousands of simulated antenna patterns. In this way, the system avoids existing interference and also uses antenna diversity to avoid creating it. In indoor environments, interference mitigation can have a bigger impact on overall performance than signal gain.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Must read: 10 new UI features coming to Windows 10