Search
Sites and Affiliations
Leadership
Research and Applications
Major Accomplishments
Frequently Asked Questions
Knowledge and Technology Transfer
Calendar
Education and Outreach
Media Resources
Technical Reports and Publications
Parallel Computing Research - Our Quarterly Newsletter
Contact Information
CRPC Home Page

UniGuide Featured Site

The Case for a Next Generation Internet
By Ken Kennedy, director, CRPC

The U.S. Congress is currently considering the fate of the Next Generation Internet (NGI), a $100-million-per- year research program proposed by the Clinton administration to fund construction of a new hardware and software network infrastructure and foster partnerships among academia, industry, and government to assure U.S. technological leadership.

The President's Advisory Commit tee on High Performance Computing and Communications (HPCC), Information Technology, and the Next Generation Internet, which I co-chair, believes that the United States stands to gain tremendously if Congress supports NGI research. Such research has the potential to revolutionize the way we do business nationwide and worldwide. Everyone, from the multinational business developing new product designs to the individual consumer making an informed purchase, will benefit from the orders-of-magnitude improvement in end-to-end communication bandwidth that will result.

So why is Congress hesitating? There are four reasons: (1) confusion over the relationship between the NGI and Internet 2, (2) concern that the government may be investing in an area whose development should be left to industry, (3) concern that the government intends to build and operate an infrastructure that will be competitive with the commercial Internet, and (4) concern that the NGI will widen the gap between the networking bandwidths available to rural states and those available in states with large urban areas.

The first of these concerns is easiest to address. Internet 2 (I2) is a consortium of universities, each of which has contributed to a pool of funds that will be used to interconnect with one another at bandwidths that are at least 100 times as large as those of the current commercial Internet (CIP 11: 1, 1997, p. 8). This goal of I2 is completely consistent with the goals of the NGI, because the federal agencies understand that the real innovations will come from the development of new ways to use this bandwidth and, historically, most of these developments have come from the research community at universities and government-funded laboratories. Therefore, it is likely that universities will be an integral component of the proposed NGI testbed.

The second concern - federal investment in an area that is currently a focus of many commercial efforts - is more complicated. Given the network speeds in the commercial backbone, can we not assume that the desired bandwidth will simply appear at our doorstep in the near future? A better question is: Will anyone be willing to pay for this bandwidth if it is made available to the consumer? The answer is that very few institutions will pay unless the benefits far outweigh the costs. To dramatically increase the benefits of high-performance networking, we are going to need new "killer applications" like those that have historically come from the research community, for example, the World Wide Web and graphical Web browsers. It is the early development of these innovative applications that will make the NGI economically feasible.

This brings us to the third concern: Should the government be building a new high-performance infrastructure that will compete with the current commercial Internet? Of course not. In fact, many of the same telecommunications companies that support the current Internet will provide the backbone services and high-speed connections for the proposed NGI. Significantly, high-speed connections can be purchased today from those telecommunications companies, yet most of the research institutions that will develop tomorrow's killer applications cannot afford to connect at those bandwidths without assistance. The connections portion of the NGI is aimed squarely at overcoming this obstacle to progress. Once the next-generation applications are available, the cost of connections will decrease through the normal operation of market forces and technological advances.

The final concern - that the NGI may widen the gap between rural and urban states - is one shared by the entire advisory committee. Networking should help to overcome "the tyranny of geography" by providing equal access to information and computing resources from everywhere in the nation. However, the cost of connection to the experimental testbed proposed for the NGI will not be uniform. Universities in states with well connected urban areas will be able to establish high-bandwidth connections at a much lower cost than states with poorer connectivity. A research initiative like the NGI can only ameliorate such problems, primarily by providing additional funds to certain disadvantaged universities to ensure high-speed access. Eliminating them entirely would require resources far greater that the $100 million per year proposed for the NGI budget.

In considering the NGI, Congress should remember that it is a research initiative, not an infrastructure program. As currently envisioned, the NGI will investigate ways to overcome impediments to the success of a new high-bandwidth global-information infrastructure that will be established over the next decade. It will create a broad-based testbed that can support research into security, reliability, quality of service, application construction environments, and new applications - all the things we need to effectively utilize the higher bandwidths that will soon be available. If successful, the NGI can forward the transition from prototype to commercial resource, ensuring U.S. leadership in this key technological area for the foreseeable future.

By Ken Kennedy. Ken Kennedy is Noah Harding professor of computer science at Rice University, Houston, TX 77005, and director of the Center for Research on Parallel Computation. E-mail: ken@cs.rice.edu