Developer News Mandrake Linux Founder Back, Virtually Amazon: We're a Technology Company Sun Expands MySQL With Closed Source

East Lansing, Michigan
Will You Pay for Bandwidth? The Berkeley INDEX Experiment

By Richard Wiggins

ou’re probably used to paying a fixed sum of money each month for your Internet access – maybe as little as $20 per month – and you find that the time it takes to paint a Web page varies dramatically depending on how congested the Net is. Would you be willing to pay a variable sum of money each month to achieve a consistent quality of service? Researchers are Berkeley are conducting an experiment to test that very question.

If you’re lucky enough to live in an area that offers cable modem or ADSL access to the Internet, you may decide to sign up for the service, figuring that your service will improve dramatically. Your provider, after all, claims the service is 50 times faster than a dialup modem. "At last, an end to the World Wide Wait!" you’ll say. Or "I can’t wait to see streaming media at 30 frames a second with a full screen image."

But those of us who’ve had cable modems for a while know better. I’ve had a cable modem in my house continuously since 1995 – running at 10 megabits per second bidirectionally – and I’ve long since learned not to expect instantaneous screen paints or television-like streaming media. In fact, the biggest advantage of high-speed local access is somewhat surprising: It’s much faster downloads of large files – e.g., you can download that 16 megabyte bundle of Netscape 4.0 in just a few minutes instead of a few hours.

In fact, some serious observers worry that massive deployment of high-speed local access will cause the meltdown of national backbones. Millions of multi-megabit pipes into homes could mean gigabits of cumulative demand on national and international links.

Some of my old friends in the computer tuning business like to say, "The science of tuning computer systems is the art of revealing bottlenecks." So it goes with the Internet: install a very fast pipe to a local ISP, and all users will suddenly share the congestion from that ISP to the backbone, and from the backbone to whatever remote servers users happen to access at any given moment.

As far back as 1993, two economists at the University of Michigan began analyzing the economics of the Internet. They saw the multimedia explosion coming, and they foresaw trouble. They put forth a bold assertion: that if a single Internet was to carry multimedia traffic with high-volume and low-latency characteristics alongside batch traffic such as e-mail, contention for the bandwidth at times of congestion would mean bottlenecks. They proposed a simple solution: people who want to move lots of data and demand that it be delivered within milliseconds should pay for the privilege.

Now one of those economists, Dr. Hal Varian, is dean of the School of Information Management and Systems at Berkeley. Dr. Varian and his colleagues at Berkeley aren’t just talking about whether consumers are willing to pay for the bandwidth they consume – they’re conducting an experiment with real users. We interviewed Dr. Varian, asking him if the world is ready to pay for quality of service (QoS) on the Internet.

More...