When deploying a wireless LAN infrastructure, organizations should follow the mantra of "measure requirements twice,...
deploy once." Otherwise, networking professionals can find themselves stuck with equipment that looks perfect on paper but fails to deliver the real-world results a deployment needs.
Rossell, administrator of network services and planning for Norwood School, a private, K-8 school in Bethesda, Md., had a serious mandate to deliver wireless LAN technology. The school was moving ahead with a plan to give all its fifth-, sixth-, seventh- and eighth-grade students wireless tablet PCs.
The school envisioned a learning environment where students could use these tablets to connect with wireless projectors to collect their teachers' lecture notes. Or students could stream their class presentations and video live to the large classroom screen.
This tablet PC deployment was an exciting project for the school but presented serious challenges that Rossell hadn't dealt with before.
"We [now] have 300 to 400 people hitting our network wirelessly all the time," he said. And in addition to a large number of devices, much of that traffic was bandwidth-intensive video, particularly to and from the projectors. And when 400 users all logged on together at 8:40, when school started, access points had a huge demand for airtime.
Rossell was initially attracted to Meru's virtual cell architecture, which differs from the way most vendors create a wireless LAN.
Instead of small, discrete wireless zones that work with the end devices on handoff, Meru uses a centralized controller to create a virtual wireless cell that spans several access points. According to Meru, this approach makes the handoff between cells transparent to endpoint devices and, ideally, reduces dropped connections.
Meru has had some high-profile customer wins in recent years, including the 41,888-seat Nationals Park in Washington, D.C., the first ballpark to have a 100% wireless coverage, according to Meru.
However, Rossell said Meru's access points simply didn't work in his environment when he went live with the students' tablet PCs.
"[Meru] is great," he said. "Their engineers are fantastic, and it all looked [as if] it was going to work beautifully. When we tested it over the summer, the video was working perfectly and projecting video to projectors, and we were watching QuickTime videos."
But when the first day of school came, the Meru wireless LAN was simply unable to handle the bursty traffic produced by students, particularly with the way Norwood's wireless projectors were set up.
"It was just with our particular application with all this video and all these projectors, we needed a solution that worked better," Rossell said.
He said that Meru came back, took down its network, worked to optimize the settings on Norwood's various end devices, and then redeployed the 30 AP wireless network, but to no avail.
Rachna Ahlawat, Meru's vice president of marketing, agreed with the generalities of Rossell's description, saying that Norwood's deployment required a special feature that they did not offer at the time, and the school was unwilling to wait for it to be rolled out.
"We had to make a strategic decision and push the features within the next 15 days, or focus on some other features," she said. "We win some other customers because other vendors won't support something. In this case, the customer would not wait, and we had to let them go."
Fortunately, Rossell said, Meru was willing to buy back its equipment. In the interim, Rossell reinstalled his legacy Cisco wireless LAN while he shopped for a new vendor.
This time, he set some new ground rules in advance: He would purchase and pay for the equipment only after it was demonstrated to work.
Aruba Networks, which has a more traditional wireless network architecture, accepted the challenge. Together, Norwood and Aruba deployed some test units and then outfitted the whole school. Overall, Rossell said, the Aruba wireless LAN has handled the students' tablet PC traffic well.
"There have been a couple of minor frustrations with Aruba in terms of their graphical user interface," he said. But dissatisfaction with the GUI is a small price to pay. "It just works, and I don't have time for it not to."
For other networking professionals trying to get it right the first time, Rossell advised playing hardball and to test, test, test.
"In retrospect, I think I would have maybe rolled out systems in parallel and maybe been less patient and loyal with the first vendors," he said. "I don't know how to save other people that kind of pain, other than think really hard about what kind of test environment you want to build."
Paul DeBeasi, an analyst with the Burton Group, said that cases like Norwood's, where the network must be installed and reinstalled multiple times, are fortunately rare.
"It sounds [as if] there was a timing issue here of the laptops and projectors all trying to access the network at the same time," DeBeasi said. "It's conceivable that with all of these laptops trying to connect at the same time, [the wireless network] would have trouble with that."
When asked about Norwood's staggering number of wireless buildups and breakdowns -- from the original Cisco to Meru, back to Cisco, a second go at Meru, and then finally Aruba -- DeBeasi was shocked.
"I never hear anything like that," he said. He concurred with Rossell's advice of hardball, best-vendor-win tactics, starting with on-paper comparisons of four or five vendors, and then narrowing down to two or three vendors in a live lab environment.
"You can do it in parallel or you can do it serially," he said. "Bring [Meru] into the classroom and run it, and then they turn off the Meru and bring in the Aruba controller, and then take those results."
For larger enterprises, DeBeasi suggested that network engineers can try purchasing a floor's worth of wireless equipment to trial before rolling it out to the entire enterprise.
This was a strategy that Norwood's (eventual) winning vendor happily encouraged.
"We tend to strongly encourage organizations to evaluate the equipment live prior to purchasing, especially large, multi-site deployments typical of K-12," said Robert Fenstermacher, Aruba's head of global education marketing. "In some cases, evaluation equipment is available for users to try-before-you-buy. In cases where an Aruba deployment doesn't meet the customer's needs, there are options to return that equipment."
Aruba does not even mind a little direct competition. "Head-to-head bakeoffs are common and strongly encouraged," Fenstermacher said.
DeBeasi said that for small deployments (the Norwood school eventually purchased 30 of Aruba's AP125s), the pre-deployment testing that Rossell described was usually enough.
"He was the unlucky one to hit something," DeBeasi said. "Networks do exhibit problems."
Fortunately, despite all the hassle, Rossell eventually installed a wireless network that supported Norwood's ambitious needs.
He also gained something else he was not counting on.
"I got really good at getting APs up and down out of the air," Rossell said. "I had [the final installation] all done in two hours because we had done it so many times before."