Quality Quest

Ensuring quality of service used to be pretty straightforward. But add in echo and data, and things get interesting.

Just how important is quality of service (QoS)? Important enough that one AT&T Wireless subscriber sued the company for signing up customers even though it knew that its network couldn't provide them with reliable service. Meanwhile, in Alabama, the governor is so fed up with spotty coverage that he backed a bill giving that state's Public Service Commission the power to regulate wireless providers based on QoS.

Even if the Alabama initiative fails, providers can't forget the argument that they made when fighting the bill: With as many as seven providers in some markets, the marketplace itself acts as a regulator, and subscribers will flee those with shoddy service.

The good news is that digital networks produce dozens of performance parameters. One barometer is handover failures.

"When the coverage is good and there's no interference, the handovers are successful very often," said Frederic Miran, Nortel Networks product-marketing manager, GSM. "When you see failures rising, you probably have a radio issue in that area."

Trending also is helpful. "We noticed that average holding time tends to get shorter on a bad radio," said Russ Arsaga, U.S. Cellular vice president, engineering. "That's because customers get upset and hang up."

Finger on the Pulse
One option is benchmarking, which collects performance statistics from multiple networks in a market. Although benchmarking provides reams of detailed statistics, much of the appeal is the score, which provides a convenient, apples-to-apples comparison even among networks using different technologies.

"We made it technology-independent so you don't have to compare frame error rate in CDMA to bit error rate in TDMA," said Don Chiang, LCC senior director, engineering. "A benchmark score is key to how the subscriber would use the network. It takes out the technology-dependent variables."

One of the more elusive problems is echo, which can't be discerned just from performance statistics. Instead, you have to go looking for it via drive testing. Some QoS tools can look at voice quality from both the wireless and wireline callers' perspectives.

"It sends sentences from the landline to the mobile and the mobile to the landline, and evaluates the voice quality along the entire path in both directions," said Tom Peters, LCC senior manager, engineering.

In the case of one CDMA provider, abysmal voice quality hinted at larger problems.

"We noticed that the voice quality on the forward link was severely degraded," Chiang said. "Initially, when there's no loading on your system, you have a fairly balanced path, and the forward-link and reverse-link voice quality aren't that much different. But in this case, we noticed a big discrepancy. We suspect that there's a problem related to congestion."

Data's New Challenges
Congestion can be even more of an issue with wireless data. That's because with both voice and data, CDMA is susceptible to a phenomenon called cell breathing, where the site's coverage area shrinks as it struggles to accommodate more users. Each new user increases the noise floor, so the site has to increase its power to communicate with the mobile and maintain the call.

In a typical conversation, pauses can account for more than half of the call's duration. But a data call lacks those pauses.

"Data subscribers have a voice-activity factor of 100%," said Jeff Miller, Safco Technologies vice president, data-collection products. "They're constantly putting energy into the spectrum, pushing up the noise floor and limiting the capacity. As the data rates increase, they push more and more energy into the spectrum."

As the noise floor increases, coverage areas shrink, and coverage holes open up.

"What you'll typically see is that the voice network has a little bit better coverage than the data network due just to data's C/I requirements," Miller said.

Poor coverage can be a double whammy in data because more packets are lost and have to be retransmitted.

"The basic difference between voice and data comes in when data networks are trying to make error-free communications and reduce the end-to-end delay," said Biju Nair, Safco Technologies vice president & general manager, software products. "With voice, you're trying to make surethat the call is sustained and there's good voice quality. End-to-end delay on a data network means the customer has to keep the call up for a longer time, and they're paying more for that."

Ensuring voice and data QoS means recognizing that they're both parts of the same whole.

"It's important for tools to be able to monitor voice and data networks at the same time whether they're planning or optimizing," Nair said.

What makes ensuring data QoS even more difficult is partnerships with third parties, such as content providers. Suppose that a subscriber complains that he can't access his stock quotes in a timely manner. The problem could be a lack of vacant channels or poor coverage, or it could be the content provider's bogged-down server. If the monitoring and diagnostics tools don't have visibility into the partner's network, ferreting out the cause could result only in dissatisfied subscribers and a lot of finger-pointing between the wireless provider and the content provider.

Add in the potential for interworking problems between protocols, and ensuring data QoS suddenly means wearing a lot of hats.

"It's just going to be a matter of time until the IP doesn't interwork well with the radio-link protocol, and you'll see large amounts of delay and lost packets," Miller said. "Somebody is going to have to dig through this mess and figure out what's going on."

One vital tool would be a network-surveillance system that can track a data call as it traverses multiple networks.

"As you move toward data services, it's going to look like a large network-management problem," Miller said. "You're going to be polling the different devices to extract performance parameters from hubs and routers versus just the RF side."

A data service might have poor performance in terms of throughput and delay, but determining why might not be possible if the diagnostic tools can't look at handsets, which affect interworking.

"Until we can extract some parameters out of there, you won't really be able to answer those types of questions," Miller said.

For wireless providers looking to data services as a way to attract and retain subscribers, QoS just got a whole lot trickier.

"We've just scratched the surface right now," said U.S. Cellular's Arsaga. "I think there will be a whole set of performance metrics developed over the next couple of years."