Just as with priority mail, some people want their Internet packets to get to there first, and they're willing to pay for such service.
But for network providers to offer priority services, the Internet needs a prioritization scheme. Researchers at the US Department of Energy say they recently achieved a milestone in creating such a system.
In an experiment conducted over the Department's own Internet backbone network, the researchers marked and delivered high-priority traffic between two of the department's labs, Lawrence Berkeley National Laboratory in California and the Argonne National Laboratory in Illinois.
"We implemented something called a 'bandwidth broker,' which processes requests for high-priority traffic," explained Stewart Loken, director of the Information and Computing Sciences Division at the Lawrence Berkeley lab.
If a user has been granted permission to use high-bandwidth applications, the bandwidth broker flags the traffic as priority traffic. At that point, the traffic can flow through the network with "just very simple processing" by network routers, Loken said. Part of that processing involves a long-time component of the Internet's Transfer Control Protocol, whose "precedence bits" were set in place with the idea of prioritization in mind, Loken said.
But the demonstration showed an overall architecture for requesting and tracking priority service without undue burden on the network infrastructure, rather than new prioritization technology.
"We're putting in place not only a workable mechanism for using the hardware features that prioritize traffic but also policy and authentication infrastructure to ensure that it's actually a workable solution," Loken said.
He contrasts it to another quality of service mechanism, the Resource Reservation Protocol (RSVP), designed to reserve bandwidth for future use. Loken said RSVP is best used among systems with a "fairly limited administrative scope."
"Reservation becomes an administrative problem when you try to scale to the scope of the Internet," Loken said. It's akin to setting up a relay of information hand-offs across the network, which is difficult to manage over the administrative structures of different networks.
RSVP is also less efficient he says, effectively holding "open" a network bandwidth pipe whether it's being used or not. In contrast, the Lab's design lets lower-priority traffic fill unused bandwidth at any time.
In the Lab's demonstration, two streams of the same video information were sent between the labs, the packets of one marked as high priority. According to the Lab's results, the first video feed was delivered at a rate of eight frames per second, while the second was delivered at a rate of one frame per second.
The test, which intentionally sent the streams on a congested path through the network, helps clear the way for more reliable and more constant Internet connectivity, the researchers say.
The Lawrence Berkeley demonstration incorporated Cisco router software to recognize specially marked packets as priority packets. The routers then shuttled such packets through before others.
While much of the Internet backbone has equipment in place that's ready to handle prioritized Net traffic, Cisco Systems' Bob Michelet says delivering service on a wider scale requires work by many parties.
"End to end service requires hardware and software at lots of points along the line," Michelet said. "ISPs and local carriers -- a lot of things work into that equation."
Loken said that's next on the agenda. In fact, a working group at the Internet Engineering Task Force is seeking priority-class traffic for use in time-sensitive, "real-time" research and business applications. The working group calls for providing quality of service in networks via a well-defined set of building blocks.
Loken says the Lab's work meets these goals and that researchers are working closely with the standards body.
"The next step is to refine this test and look at issues of extending it to the rest of the network," Loken said.
