Friday, October 30, 2009

Pandemic Would Impair Residential Broadband, GAO Says

In a serious pandemic, residential Internet access demand is likely to exceed the capacity of Internet providers’ network infrastructure, says the Government Accountability Office. That means enterprise and government disaster recovery efforts that depend on residential broadband connections may not work as planned, GAO warns.

In a serious pandemic, U.S. businesses, government agencies and schools could experience absenteeism (or forced dispersal of workers as precautionary measure) that could reach 50 percent or higher ranges, thereby displacing Internet access demand from normal daytime sites to homes, says the Government Accountability Office.

But residential broadband networks are not designed to handle this unexpected load, and could interfere with teleworkers in the securities market and other sectors, according to the Department of Homeland Security.

Oddly enough, robust network neutrality measures, such as forbidding any prioritization of bits, could render impotent one obvious way of handling the sudden explosion of traffic.

"Private Internet providers have limited ability to prioritize traffic or take other actions that could assist critical teleworkers," GAO says. "Some actions, such as reducing customers’ transmission speeds or blocking popular Web sites, could negatively impact e-commerce and require government authorization."

In other words, laws and rules that forbid "packet discrimination" would impair ability to prioritize more-important work-related uses of the residential Internet.

"Increased use of the Internet by students, teleworkers, and others during a severe pandemic is expected to create congestion in Internet access networks that serve metropolitan and other residential neighborhoods," GAO warns.

"Localities may choose to close schools and these students, confined at home, will likely look to the Internet for entertainment, including downloading or 'streaming' videos, playing online games, and engaging in potential activities that may consume large amounts of network capacity," GAO says.

"Additionally, people who are ill or are caring for sick family members will be at home and could add to Internet traffic by accessing online sites for health, news, and other information," GAO adds. "This increased and sustained recreational or other use by the general public during a pandemic outbreak will likely lead to a significant increase in traffic on residential networks."

"If theaters, sporting events, or other public gatherings are curtailed, use of the Internet for entertainment and information is likely to increase even more," GAO says. At-home workers will only compound the problem.

Oddly enough, the mechanisms ISPs could use to prioritize bandwidth so that a suddenly-scarce resource can be managed are precisely the tools strong "network neutrality" forbids.

"A provider could attempt to reduce congestion by reducing the amount of traffic that each user could send to and receive from his or her network," says GAO. "Such a reduction would require adjusting the configuration file within each customer’s modem to temporarily reduce the maximum transmission speed that that modem was capable of performing—for example, by reducing its incoming capability from 7 Mbps to 1 Mbps."

"However, according to providers we spoke with, such reductions could violate the agreed-upon levels of services for which customers have paid," GAO points out.

And that is even before any new regulations that specifically would outlaw packet shaping that could, for example, limit video streaming, gaming, and peer-to-peer and other bandwidth-intensive applications during daytime work hours, when teleworkers will have an arguably greater need to maintain functioning connections for voice and data operations essential to their work.

Overly-casual positioning of the need for "packet equality" rules can be dangerous, as the GAO points out.

No comments:

Costs of Creating Machine Learning Models is Up Sharply

With the caveat that we must be careful about making linear extrapolations into the future, training costs of state-of-the-art AI models hav...