RealTime IT News

Unidentified Fixed Wireless Events

The other morning at about 10:30 Brent Imhoff received a call on his cell phone letting him know that one sector of his small WISP network was down, again.

Imhoff is managing partner of St. Louis-based Brown Dog Networks LLC, a seven-month-old WISP with two wireless POPs, both running 5.8/5.3 GHz equipment from Trango Broadband Inc.

An hour after the call, the Brown Dog network was back upwithout any intervention by Brown Dog. Big mystery. It has happened a few times before, Imhoff says, and he fears it will happen again.

Imhoff thinks he knows what's going on. A new WISP in the area is working on setting up a link that interferes with Brown Dog's network.

Because the two networks interfere, and the newbie can't figure out how to solve the throughput problems he's seeing on his own network, he eventually shuts his radios down. End of problem. Until the next time he goes back to work on the link.

Because the outages are so brief, Brown Dog doesn't have time to deploy test equipment, so it's difficult to confirm Imhoff's theoryand impossible to pinpoint the source of the interference.

In the meantime, he admits he doesn't have a solution.

Sudden death

Sudden unexplained wireless outages are a problem, or a symptomthe underlying causes are multiplethat many WISPs have experienced, though some would rather keep quiet about it.

As Imhoff says, "It's hard to sell a business customer who needs to rely on the service when these kinds of issues are in play."

When we posted a request for comments at's ISP-Wireless list, Imhoff was one of only two WISPs willing to volunteer details of experiences they'd had.

Others suggested that reports of such problems were just rumors, perhaps started by ILECs. We think they were only joking. "It must be rumor," wrote Rudy Yakym, CEO of CyberLink International. "It does not happen here."

Others suggested with varying degrees of tact that it could only happen to WISPs who didn't know what they were doing.

There is definitely something to that, says veteran network designer and WISP trainer Jack Unger, president of Wireless InfoNet Inc.

"A lot of issues that shut down wireless networks could be avoided by proper design in the beginning," Unger says. "Most WISPs don't know this. They don't even know the limits of the equipment they're using."

Sometimes the causes of outages are just plain weird, though. We're guessing the problem that Brent Toderash of Rainy Day Software Corp. had would be difficult even for experts to predict.

Rainy Day had a link to one customer go down routinely for ten minutes every day between noon and 1 p.m. "No, it wasn't the microwave oven," Toderash writes. "It wasn't a passing train." But close.

Rainy Day techs had to sit at the client's site over a few lunch hours testing during the outage. Eventually they concluded it was factory equipment in the signal path causing interference. But Toderash admits they're still not 100-percent certain.

"Switching polarization [on the radio equipment] fixed it though," he says.

Three causes, many solutions

According to Unger, there are basically three root causes of sudden unexplained outages. Equipment failure is obviously one, but it's easily enough diagnosed and remedied.

The other big one is interference, from external sourcesor your own network. Figuring out which isn't always easy. "It can take days to analyze itby eliminating one factor after another," Unger says.

He goes back to basics to explain how interference works. Successful reception of packets by an access point depends on maintaining a low signal-to-noise ratio. The signal is the data packet someone on your network is trying to send. Everything else received at the access point is noise.

"If a competing network comes on the air suddenly and is strongly received at my access point, that's going to be solid noise," Unger explains.

"It's going to degrade the signal-to-noise ratio so that the access point won't be able to decode incoming packets. When the noise level suddenly becomes very high, throughput can drop to nothing."

Signal-to-noise is just one factor. Each data exchange on a wireless network requires a time slot. Time slots in any channel are limited.

If another network operator begins transmitting in the same frequency on the same channel, it can rob time slots from your networksometimes enough to completely block traffic to your subscribers.

If you suspect external interference, Unger says, you need to use a spectrum analyzer with a directional antenna at the wireless POP to track the source of the interference. Once you've done that, you can negotiate with the interferer to eliminate the interference.

One solution is changing the configuration of both your antennas so you're not blasting each others' access points. Another, as Toderash discovered, is to change polarization on access points, which minimizes but does not eliminate interference.

If you're using frequency hopping technology, Unger notes, it's possible to eliminate some interference by changing hopping sequences so that competing networks are no longer synchronized.

Interference effects can also occur without any external interferers, though. Unger calls it self-interference.

"If your network is on the verge of running out of time slots [because of the number of users and the amount of throughput they're demanding], and you put on a new user who hogs bandwidth, it's possible that user may dominate all remaining time slots, and throughput to existing users will drop to nothing," he explains.

There are two possible remedies for the scarce time slots problem. One is to throttle bandwidth so that no one user on your network can hog time slots. The trouble is, some network equipmentincluding the inexpensive indoor Wi-Fi equipment too many WISPs usecannot manage bandwidth in this way.

The other is to use the polling capabilities of some wireless infrastructure. In these systems, access points poll users to see if they have packets ready to transmitrather than letting users flood the network with packets. Polling makes it easier to manage the use of time slots, at least on your own network.

Again, many types of wireless equipment, including Wi-Fi, do not have polling features. So if the WISP makes the wrong choice of equipment at the outset, these problems will be more difficult to solve when the network reaches capacity, Unger says.

Viruses and worms can make the scarce time slots problem worse. They automatically broadcast packets, looking for other nodes in a network to infect. That virus/worm traffic can rob time slots from legitimate users.

If virus activity is suspected, Unger says, you need to use a protocol analyzer at the Ethernet level to analyze packets and identify non-legitimate broadcast traffic.

Typically, the packets will be emanating from one or a couple of users. To eliminate the problem, shut those users down until they can clean up.

There is one other major cause of sudden outages. When it happens, it should be a source of embarrassment to any self-respecting WISP.

The problem arises when the WISP does his site survey and testing in the winter. Come the spring, the link suddenly doesn't work. Explanation? Trees bursting into leaf block the line of sight.

"You can trace it back to lack of education," Unger says. "Ninety-nine percent of equipment used requires a wireless line of sight pathwhich involves two things."

"First, there's the visual line of sight. Then there's the fresnel zone [a clearance zone all around the optical line of sight path]. So many WISPs today either don't understand the definition as consisting of these two things, or they ignore it."

As a result, they may, often in the interests of aesthetics, place customer antennas too lowwhere spring foliage will block the frenel zone weakening the signal.

It doesn't just effect the one link either. The client device will keep trying to the send packets. So until the problem is diagnosed and remedied, the bad link can rob time slots, resulting in degraded throughput for allor even outages.

Unger is convinced most sudden wireless outages could be avoided by better design decisions when WISPs are building their networks in the first place.

Reprinted from ISP-Planet.