For the 64 percent of Americans whose internet service provider imposes a broadband cap, and for those lucky enough to have a meter, I have some bad news. The president of the firm who audits many of the country’s broadband meters says that he can’t certify the measurements produced by five out of seven of his clients’ meters because they don’t count your bits correctly.
Peter Sevcik, president of NetForecast, told GigaOM that seven clients have hired his firm to audit their broadband meters over the last few years, but of those seven only one — Comcast — has published a report on the NetForecast certification. Sevcik is only willing to certify one other client in a public report.
Meters are a black box
The other five clients — which Sevcik would not name — have meters that Sevcik views as inaccurate, although not all of them have publicly rolled out their meters. And not all of those clients impose a broadband cap. Sevcik usually expects accuracy on the meters of between plus or minus one percent, but so far these don’t measure up.
“They are wrong by missing numbers by one way or another — sometimes it’s over reporting, but more frequently the error is under reporting,” he said. Under reporting should be a relief to those facing overage charges or service termination for going over their meters, but if the meters aren’t counting the data properly, it is still a problem.
Also disturbing is the attitude that Sevcik has encountered at some clients with malfunctioning meters. “There’s a general sense by some people, ‘Eh, we under report so we give them a free pass, so why worry about that?'” Sevcik says. “I think one does need to worry because it ruins the overall veracity of the meter. It derails trust in the meter.”
Sevcik wouldn’t name those clients, but his website lists Time Warner Cable (s twc), Cox, Comcast (s cmcsa), AT&T (s T), Bell Canada, Verizon (s vz) and France Telecom as customers. Time Warner Cable and Cox have both confirmed to me that they have used NetForecast to certify their meters. While Verizon doesn’t have a broadband cap or a meter, Verizon emailed to confirm that it is not using NetForcesat on the wireline side. AT&T’s spokesman says it has a team of engineers that certifies the accuracy of its meters but that it hasn’t worked with NetForecast to certify its wireline meters. Sevcik clarified that the seven clients he’s speaking of are all U.S.-based and all are testing wireline meters.
Last November, AT&T customer Ken Stox drew attention to AT&T’s meters when he couldn’t replicate the ISP’s byte count with his own home testing. For Stox, who is technically astute, the questions he had about the meter were less about fairness and more about understanding what, when and how AT&T was counting.
Building a broadband meter is tough
Those same questions are ones that Sevcik hopes ISPs will answer as part of an overall effort to improve their meters. He notes that whatever you think about the fairness of data caps, if meters are to serve some kind of public purpose, the public has to understand what the ISPs are counting and how they are counting it.
As for problems that lead to inaccurate meters, there are several. The first is that many of these meters are bolt-on afterthoughts. A telco or a cable company often uses measurement gear that sits on the subscriber side of the network. The ISPs has to allocate enough resources at that point to track the bits properly, but networks become congested. Then the ISP faces a choice. Does it count all the bits and risk slowing down the network, or does it let the bit count slide and let the rush of packets through?
Most ISPs err on the side of letting them rush through and a better user experience. But to solve the problem they could dedicate more resources to the counters so they can keep up with peak traffic. More resources would also solve the next problem ISPs face — once they have the bit counts, they need to add them up. As Sevcik describes it, many of these counters drop the bits into an Internet Protocol Detail Report format. Those reports are generated every 15 minutes.
Spread that across 10 million subscribers with a goal of doing hourly updates, and suddenly you have 40 million records to process in that hour. That takes servers — in some cases more than the ISP anticipated.
And while Sevcik said that while some ISPs had used decimal counting as opposed to binary counting of bytes in the past, most used binary counting today. That’s good because a binary count adds about 7 percent to the total number of bytes. But as Sevcik points out, if a consumer streams 3 HD movies on Saturday night and expects to see that jump in data consumed on his usage meter, then it needs to be there, or the consumer needs to know why.
Are meters worth it?
It’s clear that building a meter takes work. A Time Warner Cable spokesman notes that development of its meter took several years — other ISPs said it took at least a year of effort from multiple engineering teams.
If building a meter is so much work and consumes so many resources, why have them? For example, Comcast, which delayed the rollout of its meters while it struggled to get it right (and still runs monthly accuracy checks) has defended its meter as a customer education tool and as a means to manage network consumption. Critics point out that at 300 GB, Comcast’s cap is suspiciously close to the 288 GB figure that Comcast has said would be the amount of data consumed by someone using their broadband to replace cable. Those critics generally call caps a way for ISPs to protect their pay TV businesses.
What we do know is that Comcast has spent a lot of money and effort making sure its meter is accurate, because as Charlie Douglas, a Comcast spokesman notes, “We knew it would be in the spotlight.” I imagine also because it knew a meter would be the first step in how it could change the pricing dynamic from all-you-can-eat to something that’s a little bit more metered. And as I have pointed out in previous articles, if meters become the basis for charging subscribers overage fees or even terminating their service, then someone needs to monitor those meters to ensure that they are accurate.
I’ve called on the FCC to wake up and start gathering more data on how meters affect consumers and whether or not they are accurate, but the agency has so far been content to let this experiment in caps pay out without much oversight. With these accusations maybe the FCC will finally step up. Clearly, as a country we’re moving toward capped and metered broadband.
Sevcik, whose experience goes back to the days of the ARPANET and the first routing systems, believes if that’s the case, then those meters should be accurate.
“I’ve been in the internet business for quite some time … and in that time I’ve had my hand in the design of more than 100 networks and seen a lot in network technology. And what I’ve realized is, as the industry has matured there is an awful lot of talk and decisions made by people — consumers, policy advocates in DC and big companies — that is often based on hype,” he said. “my goal in a small way in this world of hype is to shed a light of real data and make a little piece of it really right.”
In short; If we’re going to accept meters on our broadband, then let’s make sure they are accurate.
This story was updated at 11:49 a.m. PT with comment from Verizon.