BERLIN — State-of-the-art Web surfing, for all of its breathtaking speed, can be baffling. A favorite page gets hung up. A data-intensive application, like playing a video or downloading large files, stutters or stops. Is it the telecommunications operator? Is it the Web site? Is it the smartphone or the computer? Or just a sign of Internet thrombosis?
Krishna Gummadi, the head of the Networked Systems Research Group at the Max Planck Institute for Software Systems, in Saarbrücken, Germany, says the blame often lies with the telecom operator, which is selectively slowing broadband speeds to keep traffic flowing on its network, using a sorting technique called throttling.
In 2008, Mr. Gummadi and a graduate student, Marcel Dischinger, developed a free software gauge that detected whether broadband service was being throttled by a network operator. The software, called Glasnost after the Russian word for “openness,” has been downloaded and used by 1.5 million people around the world since then.
Glasnost mimics data transfers using the Bit Torrent file-sharing protocol, and then measures whether operators are slowing uploads and downloads. Consumers around the world have used it to test the service of landline broadband operators. Glasnost only works on a few smartphones so far.
The latest results, based on 121,247 tests run from January through October, suggest that throttling is being done everywhere in the world. The results for each operator may not be representative, for several reasons. The sample sizes for each operator vary, from 36,000 in the case of NTT Docomo of Japan, to just a hundred for smaller ones. There is also a 4 percent to 5 percent chance of “false positives” — indications that throttling is being done when it is not.
But that aside, Glasnost cast light on the practice of throttling, which operators are reluctant to discuss in detail beyond confirming, in standard service contracts, that they use it.
In the United States, throttling was detected in 23 percent of tests on telecom and cable-
television broadband networks, less than the global average of 32 percent. The U.S. operators with higher levels of detected throttling included Insight Communications, a cable-television operator in New York, Kentucky, Indiana and Ohio, where throttling was detected in 38 percent of tests; and Clearwire Communications, where throttling was detected in 35 percent of the tests.
Throttling was detected in 18 percent of tests on Verizon’s landline network and in 30 percent of tests run on AT&T WorldNet Services, the company’s consumer broadband network. Throttling on AT&T’s business network, SBIS-AS AT&T Internet Service, was 18 percent.
In Europe, throttling appeared to be most common in Britain. Slowing was detected on 74 percent of tests done on BT’s British regional network. Positive tests for throttling also exceeded 50 percent for six other British operators: NTL, Opal Telecom, Telewest Broadband, Carphone Warehouse Broadband Service, Tiscali U.K. and Pipex.
In France, throttling appeared to be less common. Positive tests didn’t exceed 21 percent among France Télécom’s Orange service, Neuf Cegetel, Numericable and Proxad. In Germany, it was even rarer, at levels of less than 16 percent for almost every operator including Deutsche Telekom. (I tested Glasnost on my Deutsche Telekom network in Berlin and it showed no throttling.) The one exception: Kabel Deutschland, the biggest’s domestic cable TV operator, showed throttling detected in 44 percent of 393 tests.
In Japan, NTT Docomo employed throttling in 49 percent of 471 tests, according to Glasnost. GigaInfra Broadband and Vectant had positive tests in 30 percent and 38 percent of tests, respectively. In Canada, where the population is much more spread out, and networks must cover vast territory, throttling appeared more common. It was
measured in 85 percent of tests on Rogers Communications’ network and 64 percent of tests on Bell Canada.
In other parts of the world, frequent throttling was detected in smaller operators, which often have less money to build high-capacity networks. Those included: the Dubai-based Emirates Integrated Telecommunications, operator of the Du network, with 90 percent; Toya, a cable operator in Lodz, Poland, with 88 percent; TeleCentro of Argentina, with 87 percent; RLE Elisa in Estonia, with 85 percent; ASN AtHome, a Hong Kong-based cable TV operator, with 83 percent; TM Net of Malaysia, with 78 percent; Magix of Singapore, 63 percent; Cabo TVM of Portugal, 62 percent; and Bezeq of Israel, 59 percent.
Former monopolies like Telefónica of Spain, Telecom Italia, KPN of Netherlands, Telstra of Australia, Telia of Sweden, Belgacom of Belgium and Eircom of Ireland, which all still operate the largest landline networks in their countries, generally used throttling less frequently — perhaps because they didn’t have to, on their extensive networks.
Their rates of detected throttling, respectively, according to Glasnost, were: 19 percent for Telefónica and Telecom Italia, 18 percent for KPN, 34 percent for Telstra, 14 percent for Telia, 13 percent for Belgacom and 15 percent for Eircom.
In general, the Glasnost results suggest that telecom and cable TV operators, when they do use throttling, do so mostly to suppress bandwidth hogs and ensure a reasonable experience for all of their customers. Mr. Dischinger, now a computer engineer in Innsbruck, Austria, said throttling was much more commonly used by operators of mobile phone networks, which have much less capacity than landline grids.
But with operators starting to sell superfast landline broadband service for heavy data users, such as Deutsche Telekom’s high-speed fiber-to-the-home service, the competition for bandwidth — and the need for throttling — will only increase, Mr. Dischinger said.
“I highly doubt it can go on forever,” Mr. Dischinger said. “I cannot envision with the current network infrastructure they have that operators can continue to support people in the long term without more investment.”Read more at www.nytimes.com
- prakaas89 posted this