Network Computingwww.networkcomputing.com

The Drive for Automation
Users are also taking advantage of autoconfiguration for routers to use secure VPNs, which in turn is facilitating automation of business-to-business e-commerce. In addition, IPv6's routing features support the use of multiple IP addresses on a single physical address, which means that instead of spending an inordinate amount of time tweaking NAT and DNS entries, you can simultaneously deploy and support IP addresses for your VPN.

The key point to remember about the Internet as it stands is that nothing is turnkey. The architectural developments of the Internet itself, as well as the tools you need to conduct online commerce, are coming on fast, to be sure. But your imagination is not yet the limit when it comes to applications; the limit is determined by the price you can afford for retooling and workarounds. Next year, however, may tell a different story.

Rivka Tadjer is a freelance journalist and technology columnist based in New York. She writes a column on electronic commerce for The Wall Street Journal's interactive edition. Please send your comments on this article to her at rivka@reportersink.com.

Internet2: Testing 1, 2, 3, Testing
To hear some people talk about it, the Internet2 sounds like a disjointed network, a new Internet. But it's not the Internet reinvented; rather, it's a microcosmic mirror of the Internet's current architecture, designed to provide a research environment.

The idea, according to two of its biggest sponsors, Bay Networks and Cisco Systems, is to use it to test things like multicasting. If something goes terribly wrong during a test on a separate network that is simply a scaled-down version of the Internet, it won't affect the public space that the Internet has become. Also, if some new feature is developed on a mini-Internet that is architecturally the same, it will be very simple to hoist it onto the Internet.

That said, the three priorities for Internet2 developers these days are augmentations, busy signals and multicasting. Here's the latest word on each:

·Augmentations: The idea here is to create a scheme to differentiate between voice and collaborative video traffic. To that end, developers are using boxes called bandwidth brokers. Network managers talk to them via a Java interface and express a traffic priority policy. So, for instance, voice flowing through the network is treated with a different priority than video. The broker box talks to the routers, switches and other equipment that the IP packets will go through; then the packets come in and are marked by the edge router, which labels video and voice differently.

·Busy signals: The QoS (quality of service)-enabled network is fighting back against network congestion, spurring a paradigm shift for applications developers, users and network planners.

There are essentially two ways the network can respond when congestion occurs. New connections can be admitted and everyone experiences degraded performance, which is, of course, characteristic of today's Internet and of precedence-based differentiated services schemes. Alternatively, new connections can be refused and the quality of existing connections maintained, which is characteristic of the phone system. It is also becoming characteristic of QoS approaches that offer nonrelative transmission guarantees.

For many applications, it may be better to experience an occasional busy signal during call setup than to face unexpected degradation of service. For example, researchers visualizing a computational simulation on a supercomputer using a distributed collaborative visualization tool cannot afford to have their technical debate cut short because of unexpected congestion on the network. By explicitly reserving resources end-to-end during call setup, the architecture can support a busy signal when the required network resources cannot be guaranteed.

·Multicasting: There has been more noise about multicasting than any other Internet2 testing. The idea is this: If you have a system that sends out data and sends it out only once, the system shouldn't need to know its specific receivers, only that many systems receive the data. Multicasting is one-to-many broadcast, like radio; end users simply tune to a station. With the Internet, users may tune to a virtual channel, which could be a link from a Web site. Multicasting is now used at the New York Stock Exchange for live market feeds.

The Internet's developers are also focused on providing QoS for multicasting because the Internet's major drawback continues to be inadequate bandwidth. The decisions they're debating now are whether to use the lowest common denominator for bandwidth--since some places on the Internet have high bandwidth and some have low--and, naturally, how to make it scale. Even the stock exchange broadcasts to thousands of systems, not millions, so scaling is really untested.

The challenge is ironing out the algorithms that handle the multicast trees and determining how to route them and how to provide QoS for time-sensitive data. The assumption is that the Internet will handle all phone calls, video and commerce.

So far, Internet2 developers have a general sense of what the architecture looks like. But some pieces, such as security issues particular to multicasting, will take time to gel. Say someone calls you every three seconds to conduct telemarketing or just to be annoying; that's enormous resource consumption if it's a commonplace occurrence. That's not new, and neither is a trojan horse that can steal data anonymously. But new or not, they are threats that still pose problems.