LibriumArcana.Games

arstechnica.com wrote:Wireless carriers like to say that monthly data caps are necessary to prevent heavy users from slowing down less active ones.

After surveying the four biggest carriers this year, the US Government Accountability Office reported that “some wireless ISPs told us they use UBP [usage-based pricing, i.e. data caps] to manage congestion.” Verizon Wireless has insisted to the Federal Communications Commission (FCC) that data caps are so effective at reducing congestion that they eliminate the need to throttle most customers.

I won't argue that data caps have no positive impact on wireless networks—they can prevent the most egregious overuse of what is a limited resource. But it's a crude tool at best, targeting monthly averages with no regard for whether the network is congested at a particular time or place.

Recent actions and statements from carriers suggest this is the case: data caps are largely a profit play, not an efficient means of preventing network problems. After Sprint offered “double the high-speed data" on its network, 20GB per month for family plans, AT&T responded by doubling data, too, through shared plans of 30GB to 100GB a month. Verizon doubled its own customers’ data, while Sprint offered yet another doubling to stay ahead of AT&T and Verizon. Suddenly, network constraints had apparently disappeared.

Where did all this extra capacity come from? The carriers’ networks didn’t double in size overnight. The capacity was always there—carriers just weren’t allowing customers to use it until one decided to boost data and the others followed. Behavior like this helps explain why federal regulators have blocked mergers that would reduce the number of nationwide carriers from four to three.

Data caps aren't about protecting the network

If you’re driving in the middle of nowhere and need cellular data to get directions, that data usage isn’t going to slow down anyone else on the network. Backing your phone up to the cloud in the middle of the night probably won’t affect other subscribers, either. Yet it all counts against your monthly data cap, the carriers’ most obvious means of ensuring that networks don’t get overloaded.

“If you try to use monthly volumes as a way of managing congestion, you never do a very good job of it,” Sandvine cofounder and CTO Don Bowman told Ars. Sandvine is known for its research on Internet traffic and its products that help Internet providers (including Comcast) manage network congestion.

Data caps “are more about defining a differentiated value than they are about protecting the network or protecting the other users on the network,” Bowman said. “You don’t want to launch a lower price point and have your higher users jump down into it.”

It probably wouldn’t be smart for carriers to promise everyone limitless data, because there are real capacity constraints in wireless networks. But the specific limits imposed on consumers are chosen not to prevent congestion but to maximize profit.

If everyone had unlimited data, they might “use as much of it as they can because it doesn’t cost them anything, and the whole thing collapses for everybody,” Bowman said. “There is a value in having the limit there, but it’s not about individual locations being busy. It’s about the network capacity as a whole on average. Usage management is a way to align the average cost of running the network with the average revenue you make per user. It’s not about the peaks and valleys.”

Unlimited data isn’t completely dead. Sprint, which still offers such plans, acknowledges that congestion management and monthly data limits “are two separate issues.” T-Mobile offers unlimited data but throttles users after they hit monthly limits on high-speed data. But AT&T and Verizon, which have the two best performing networks, abandoned unlimited data for new customers and are trying to push longstanding customers from unlimited data to capped plans.

Data caps help carriers make money in all sorts of ways. Most obviously, carriers tack extra charges onto customers’ bills when they exceed monthly limits, which may convince customers to either reduce their usage or upgrade to a pricier plan. Carriers can also create a two-sided market by charging content providers for the right to offer services without counting against the customers’ caps. (AT&T has pioneered this approach in the US.)

Data caps are also easier to implement than management tools that detect congestion at specific cell sites and throttle the heaviest users to ensure a smooth experience for everyone else.

AT&T, Verizon, Sprint, and T-Mobile all say they have the ability to selectively throttle users in times and places of congestion. But instead of throttling solely to prevent network overload, they’ve designed throttling policies to affect only certain users that they want to push onto pricier plans. Even former wireless industry lobbyist Tom Wheeler, now the FCC chairman, says it’s true.

Wheeler and others have objected to carriers throttling only those users who had been grandfathered into unlimited data plans that are no longer offered to new customers. When ordered to justify its throttling policies, Verizon Wireless told the FCC that users without monthly limits are the only ones who need to be throttled because their plans don’t provide incentives for limiting use.

In reality, Verizon and other carriers let customers on limited plans use far more unthrottled data than “unlimited” customers as long as they pay overage charges or buy more expensive plans. Rather than throttling solely to prevent congestion, carriers throttle to push customers onto plans with data caps.

How carriers describe their throttling programs to the FCC matters because they’re also trying to convince the commission that they shouldn’t be regulated like fixed broadband providers. "Mobile broadband faces unique operational constraints," they say. After FCC pressure, Verizon reluctantly agreed not to throttle 4G users, but it still throttles those on 3G.

Network management strategies are “coarse”

Carriers’ network management methods probably aren’t as sophisticated as they would have you believe. Ideally, they would evaluate traffic in real time at individual cell sites and only slow down users who are hogging bandwidth at that exact moment.

“It’s actually quite technically challenging to get very accurate with that,” Bowman said. Sandvine sells tools that do this real-time management, but “it’s not what the big four are using today,” he said. “Their strategies are much more coarse.”

Under ideal circumstances, LTE networks can provide 4 bits of throughput per second for each hertz of spectrum, Bowman said. With a healthy 20MHz of spectrum, that's 80Mbps that could be devoted to areas of around 250 meters (or even smaller) by using small cells, Bowman said. “That’s a lot of data in a farmer’s field, and it’s often not a lot of data in a New York subway system,” he said.

The carriers don’t say exactly how they choose where and when to throttle, but they provide hints. AT&T, Verizon, T-Mobile, and Sprint all say they throttle only the top 3 to 5 percent of users, as measured by monthly data consumption. Once you use a certain amount of data each month, you’re at risk of being throttled if you enter a congested area. That means carriers are throttling customers based on their past usage without considering whether those customers are negatively impacting other users at the exact moment they’re being throttled.

As for deciding which cell sites are congested enough to merit throttling heavy users, carriers may also be basing that upon historical usage patterns rather than real-time evaluations, Bowman said. Real-time management is more expensive and “more complex to implement than a simple billing cap type of approach,” he said.

There are alternatives to monthly data caps. “There’s a reasonable number of mobile carriers in the world who have time of day pricing on bytes,” Bowman said. “You’ll see that with satellite companies. They often offer unlimited volume off-peak and limited volume on-peak. One of the problems that causes is it can chase the peaks around.”

Data caps do have the virtue of being simple to understand. Still, there might be a fairer method of billing that involves varying charges for data packets based upon congestion, wireless consultant and engineer Steven Crowley told Ars.

“It would take more complex policy and charging software in the operator's network,” Crowley wrote in an e-mail. “How this would work on the user side needs some thought. How is the user made aware of the instant charge for a data packet? It has to be convenient and transparent. In the Washington, DC, area we have toll roads that vary the toll dynamically according to traffic, but there's a big sign saying what the toll is—then I get to decide if I want to use it. Data usage is becoming more transparent and easier to monitor; that evolution might lead to a satisfactory peak pricing experience for the average user.”

FCC talks about caps, hasn't taken action

Wheeler forced Verizon's hand on 4G throttling after making it clear that he objects to carriers basing network management decisions on customers' data plans rather than on network architecture and technology. He may have other opportunities to take action. T-Mobile, as we’ve reported, accused AT&T and Verizon of charging “monopoly rents” for data roaming, making it difficult for smaller carriers to offer unlimited data or plans with higher caps. T-Mobile asked the FCC to step in and force AT&T and Verizon to negotiate lower rates.

Separately, US Rep. Anna Eshoo (D-CA) is urging the FCC to examine data caps on both fixed and wireless networks as part of its network neutrality rule-making process. Wheeler wrote back to Eshoo, noting that he is concerned about wireless throttling and that the FCC is gathering information about how broadband providers "use data caps to manage traffic, and whether they can be used to exploit market power or reduce competition."

The FCC's tentative net neutrality proposal from May of this year did not target caps and imposed weaker rules on wireless service than fixed-line providers, anyway. The FCC could alter the proposal, but for better or worse, data caps are probably here to stay.