Wired Networks Face Limited Future

Xirrus CEO Dirk Gates sees major growth in WLANs as enterprises deploy bigger and faster wireless networks to meet demand from pervasive mobility and consumerization of IT.

May 16, 2011

15 Min Read
Network Computing logo

Dirk Gates, CEO of Xirrus, the wireless networking company, sat down with Jonathan Feldman and Lorna Garey at Interop 2011 in Las Vegas, a UBM TechWeb event, to discuss his company's wins at Carnegie-Mellon University and the Port of Houston, as well as the future of wired Ethernet.

InformationWeek: Some IT pros have concerns over the very limited range of 802.11a, especially in buildings, and the efficacy of beamforming techniques. An iPhone, for example, doesn't have a directional array, and it can't communicate back through all that concrete.

Dirk Gates: Actually, it can. Think about a megaphone. If I put a megaphone up to my mouth and I'm shouting, you can hear me at a great distance. If I turned it around and put it up to my ear, I can hear you at great distance. So the directional antenna has gained both in transmit and receive path, so it amplifies my transmit signal, but it also amplifies the incoming receive signal. So I get a balanced link and balanced gain in both directions. I am able to maintain a conversation with you, with that directional antenna twice as far away as from our device as you would from a standard omnidirectional antenna on an AP [access point].

InformationWeek: So what you're saying is that instead of doing omni to omni, you're doing directional to omni?

Gates: Right, in both directions. So I get the advantage on both sides; that allows us to have a quarter as many devices provide the same sort of coverage as traditional APs. I have just as many radios, but they're packed into one device with long-range directional antennas instead of having to hang four devices to get the same coverage.

InformationWeek: Tell me a little bit about your field experience with this. Can you give me some examples of older buildings that might [be] Wi-Fi hostile?

Gates: A great example is at Carnegie-Mellon University. They were early adopters of .11 technology. In the late '90s, they put in 1,000 Lucent .11b access points, all of them 2.4-gigahertz only. They had coverage throughout all these dorms, with a lot of cinderblock.

But recently [Carnegie-Mellon was] looking to upgrade to .11n and 5 gigahertz, just to handle the capacity. And they correctly assumed they would need twice as many APs to be able to operate at 5 gigahertz, because the range would be less. So they were looking at going from 1,000 Lucent .11b APs to 2,000 dual-band .11n APs. And we were able to come in, do an active site survey, and actually demonstrate to them that we could cover that same area--all those dorms--with 300 of our eight-radio devices. I still have 2,400 radios in the air. But because of the directional antennas and the nature of the product, we were able to get complete 5-gigahertz coverage with just those 300 devices.

InformationWeek: What's the price point per radio that you're looking at, and what was the return on investment?

Gates: About $700 a radio. Actually, probably a little bit less. It's close to $700 at list.

InformationWeek: And you pack anywhere from four to eight to 16 in one array device?

Gates: When you take a look at [Carnegie-Mellon's] return on investment, their original plan is they were going to roll out 2,000 APs over 12 months, about 500 in a quarter. They would have to pull 1,000 new cable drops--and at union labor in Pittsburgh, that was running $750 a cable pull--and then they would have to take the 1,000 100-megabit Ethernet switch ports and put in 2,000 Gigabit Ethernet switch ports. So they were looking at $1 million.

InformationWeek: And to be clear, that's because they wanted throughput of greater than 100 Mbps at each one of these APs?

Gates: On a bonded channel you can get close to 200-plus megabits on one radio and two radios. So you're going to need more than 100 megabits ... They're going to put in 2,000 Gigabit Ethernet switch ports. We were able to cherry pick 300 of the existing 1,000 cable pulls. No new cable. Instant $750,000 savings. We were able to do this with 300 Gigabit Ethernet switch ports, not 2,000. Hundreds of thousands of dollars in savings here as well.

InformationWeek: Can you use standard PoE [power over Ethernet]?

Gates: We have our own PoE solutions. The device is large enough that it needs high-powered PoE. So they did buy rack-mounted PoE solutions from us. The alternative would have been 2,000 PoE Gigabit Ethernet ports. They did 300 PoE [ports] using our equipment.

POE today, standards-based, is less than 15 watts, and you're not going to run eight or 16 access points on 15 watts. There's a high-power PoE standard in the works, and it will get you to about 30 watts.

InformationWeek: And what do you guys drop?

Gates: About 50, 60 watts on the large arrays.

InformationWeek: You hate the Earth, don't you? [Laughter]

Gates: Well, using our PoE solution, which is SNMP-controllable, you can turn them off at night. It's programmed to go off and come back on.So in any case, $1 million in cable savings and switch port savings right out of the gate, and we were able to put in the 300 devices in 60 days. They were looking at rolling 500 a quarter. We did the 300 in two months. So in the span of July/August, before the school year started, the network goes in, up, running. So how do you calculate ROI? Well, the million-dollar savings is a given, and there's the fact that you had an operational network in two months instead of 12. All the students had advantage of the network for the full school year.

InformationWeek: And the reason that it would have been delayed was because of union cable drops or just because between IT's labor and contractors' labor, that was sort of the work capacity that they had. Why?

Gates: They just figured that they could put 500 access points up per quarter. So whatever that boils down to--125 or 130 a month.

So you figure they're hanging up six access points a day. They're pulling a cable, putting in an access point. So when you look at that, 2,000 devices will take them a year. Three hundred devices, two months. Huge.

When you take a look at deploying very scalable, high-capacity wireless networks that can handle that sort of device count, this becomes a much more economical way to do it.

Interestingly enough, at CMU, when we installed the network, that first fall when the kids came in, there are 5,000 undergraduate students at the school and probably 4,000 in the dorms, and there were 4,000 unique MAC addresses. Makes sense. Beginning of the school year this year: 7,000. I went and visited just a couple weeks ago: 13,000. That's an iPad, a phone, and in the dorms, gaming consoles, printers.

InformationWeek: No wonder we need IPv6, right?

Gates: That's been our biggest challenge. They had planned their DHCP pools to handle one address per student. Beginning of the school year we got a frantic call: The wireless network isn't working. So we sent our guy out on site. We helped them reconfigure their DHCP pool to be larger. That's when they discovered they had twice as many wireless devices [as they'd expected]. And so now they're aware of it, and they're thinking about larger pools. But the average is now three devices per student.

Another interesting thing they did: Beginning of this school year, they turned off all of the Ethernet ports in the dorms. They noticed that they weren't being used anyway. Just shut them all down. And you now have to go to the IT help desk and fill out a port-activation form [to get a wired port] and I think the number's less than 100. So students that absolutely want to be on the wire have to fill out a port activation. But less than 100 of those 4,000 students have.

InformationWeek: And how's that benefiting CMU?

Gates: First of all, they don't have to manage them. And in many cases they've just turned the switches off.

InformationWeek: And maybe they can recoup some of those switches, reuse them. Is that the idea?

Gates: That or eventually they're going to decommission them and not have to pay maintenance on them.

InformationWeek: That's a really great example. Do you have another industry or enterprise example?

Gates: Port of Houston is a good example. Port of Houston, government agency, wanted to use Wi-Fi in the docks. And they were looking at hanging I don't know how many hundreds of Cisco APs on all these light poles around the dock to be able to provide the Wi-Fi coverage, and then figuring out how to backhaul all of them. We were able to cover the entire port with 40 devices. They're hanging up on the light poles. They're about 100 feet in the air. We have enough radios in each device that [they could use] a couple of the radios for wireless backhaul between the arrays.

So out of those 40 devices, less than 10 have got a fiber backhaul into their network. The rest of them are connected with these wireless distribution connections, array to array. So we have less than 10 fiber links and only 40 devices hung, we got pervasive Wi-Fi coverage on the entire dock. This would have been hundreds of access points--and very likely every single one of them needing backhaul to make that work.

In the manufacturing scenario as well, because of the long range of the device, we covered a much larger area, whereas to hang individual access points I would need to put closets in the middle of the floor. I've only got 100 meters from any one direction, and so building an IDF [intermediate distribution frame] from the middle of the floor is not an attractive solution. [We enabled them to] eliminate that and cover the entire space with fewer devices and fewer poles.InformationWeek: One big theme at Interop this year is consumerization. When you talk to CIOs, how do you see that playing in to their Wi-Fi strategies?

Gates: Absolutely. That is a big attitudinal shift that we're seeing. In the past it's absolutely been about, "I'm locking down my network. I don't want to have Wi-Fi at all. If I must, I'll put it in the conference rooms, but only under duress, and if the CEO wants it I'm putting an access point in his office, but no one else."

Then with iPads and these devices that don't have RJ45s in them, people are bringing their own devices. I've got an iPad, I'm bringing it to the office. I've got my own smartphone, I'm bringing it to the office. Why is it that at home I can have coverage and Wi-Fi everywhere, but I can't get it in the office?

There's this almost revolt occurring, and you're seeing this shift in attitude where they see the tsunami, and you talk to these guys and they finally realize, "OK, I've got to figure out how to support the whole bring-your-own-device movement. I've got to make iPads workable anywhere." The difference in the past is you'd pick up your notebook and you'd go to a conference room. At your desk, you're on the wire. Conference room, wireless, but I don't need wireless anywhere else. Now with an iPad, I need wireless at my desk.

InformationWeek: All those students are coming out of college, too.

Gates: All those kids at Carnegie-Mellon, with their average of three Wi-Fi devices. As they move into the workforce, they're part of that tsunami, they're part of that revolt. So I think we are seeing more interest in our solutions from folks that have figured out they have to embrace the bring-your-own-device, they have to go from hotspots and conference rooms to pervasive Wi-Fi throughout their facility. And by the way, they don't want to bankrupt themselves by pulling millions of feet of cable and adding even more Ethernet switch ports to their infrastructure when they already know that half of them aren't used. It's a frustrating thing. They just want to pull out those old RJ45s and reuse them for APs. That attitudinal shift in the enterprise to pervasive Wi-Fi is something we're picking up on. People are saying, "OK, I've got to figure out how to do this, do it cost effectively, but I gotta make sure that those iPads work at your desk.

InformationWeek: What do you think are the chances of organizations going from wired at every work cube to going completely wireless, pervasively? And here's where it's not a softball question: Tell me why.

Gates: I don't think the wire goes away.

InformationWeek: You don't?

Gates: They're going to like ice cream, but you know, they like cake as well. But the difference is there will be less cake. So right now, [in most companies,] I'm going to venture to guess that half or two-thirds of those Ethernet taps aren't used. The constant mindset for Ethernet deployment has been over-provision. As long as I'm putting this stuff in, I'm going to pull extra cables. Why just one RJ45 to your cube when I can put in three? 'Cause that's going to go to a phone, and maybe you'll need something else.

So there's been this massive over-provision and this inflated market. That I think gets right-sized. I think that in the future we're not going to be pulling extra cable to your cube, one tap will do, you don't need three, because I've got the Wi-Fi network. So you're going to see this shift toward right-sizing the provisioning of wire and starting to go from an under-provisioned Wi-Fi approach, where people use it just in the conference rooms, or maybe the barest, thinnest of coverage in the building, to starting to think about over-provisioning Wi-Fi. I'm going to put up more radios than I potentially need today but knowing that I'm going to have increased use in the future.

InformationWeek: The benefit to the organization is?

Gates: That now I can support you bringing your own device. In the future, I don't have to be buying notebooks for everybody. I don't have to be buying iPads. I'll let you pick your own. By the way, pay for it yourself. And I can let you bring your own phone in. The phone's already gone that way. You go back a decade ago, people were supplying phones as part of the business. That doesn't even happen today. "I'm going to pick my own phone, it's going to be what I want to use, and I'm going to use it, and if you don't help me I'm going to figure out a way around you." And notebooks are sort of trending that way, and I'm sure that companies aren't giving away the best notebooks, so people are just saying, "I don't want your standard issue; I'm going to bring in my own." And iPads have just blown the thing wide open.

So, if they embrace bring-your-own-device, it can actually make their life easier in terms of not having to enforce that standardization in a militaristic sort of way. It can save them money. And, it changes the whole daily workflow for the IT guys. They're not configuring machines and handing them out, saying, "Don't change it." They're going to have to figure out how to support all the things that come in. There is a cost savings to the business, but it's a different day for the IT guy.

And, I've got two boys in college who don't know what an RJ11 is. They've never had to dial in with an analog modem. I've got two little ones that are in fourth and fifth grade as well, and I'm going to tell you that by the time they reach college, they're not going to know what an RJ45 is. They've never used an Ethernet tap in their life. The boys used to, but these two little ones, give them another five or 10 years, and they're going to be out there, like, what's an RJ45?

InformationWeek: One last question. What's your take on the swing from dumb controller to smart controller, dumb AP to smart AP?

Gates: We used to in Wi-Fi have the luxury, because we had very low speeds, of pulling all the traffic back to a controller, decrypting, doing our encryption-decryption there, doing all of our traffic-flow management in a centralized controller. And when you've got your top data rates of 54 megabits, you can do that for quite a few devices.

Now that the data rates have moved to 300 megabits, and they're moving to 450 megabits, and with the standards on the horizon you're going to be at Gigabit-plus speeds in the air, trying to funnel all that traffic back to a central controller becomes very untenable. You're going to have to have tens or hundreds of gigabits of throughput in that controller.

So just as Ethernet moved the intelligence out to the edge so as soon as it comes off the client devices, you handle it and take care of it, intelligence is going to be distributed and moved right back out to the edge in Wi-Fi. You're going to have to be able to handle that sort of traffic. We started that way from day one. We put controllers in each of our devices at the edge, we do all of the filtering and encryption-decryption, everything right there so you don't have to worry about bottlenecking at a controller.

Recommended Reading:


Vendors are fighting it out in the market for integrated network, computer, and storage systems. In the new all-digital issue of Network Computing, we go ringside to help you pick a winner. Download the issue now. (Free with registration.)

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
More Insights