NACSE - Northwest Alliance for Computation Science and Engineering
NERO - Network for Education and Research in Oregon

Testimony of
Cherri M. Pancake
Professor of Computer Science and Intel Faculty Fellow
Oregon State University

before the
Subcommittee on Communications
Committee on Commerce, Science, and Transportation
United States Senate

June 3, 1997

Introduction

Mr. Chairman and distinguished members of the Subcommittee, my name is Cherri Pancake. As Professor of Computer Science and Intel Faculty Fellow at Oregon State University, I pursue research in the general area of high- performance computing and communications. I am a member of the National Science Foundation's advisory panel on computing, information, and communications, and also serve on a number of advisory panels for the high-tech industry, national laboratories, and academic consortia.

The Contribution of "Middle-Class" Universities

Today, I'm representing Oregon State University in its character as one of what I will refer to as the nation's "middle-class" universities. Among universities, the top-50 schools (such as Stanford, MIT, or Harvard) function much like an "upper-class." They are the most prestigious institutions, benefiting from very sizable endowments and receiving a large proportion of external funding from both government and private sources. In contrast, the institutions whose perspective I will describe are ranked 51-150 nationally. They're really a very different class in several respects.

First, almost all of us are public institutions. We don't have large endowments or other resources we can call upon in starting new initiatives. Because of our lower ranking, we have to work extra hard in order to attract research dollars, faculty, or students. But middle-class universities do, in fact, contribute significantly to national research, particularly in applied fields such as agriculture, engineering, and natural resource management. For example, Oregon State ranks 60th in terms of overall research funding, but leads the nation in both federal and non-federal research grants in the field of forestry. In fact, 43% of the universities that have been conferred "Research I" status (by the Carnegie Institution) are in the middle-class.

We also make some unique contributions to the national educational infrastructure. Consider the nation's land-grant institutions, of which only 19 fall into the upper-class category. Land grant schools play the central role in extension and lifelong learning - touching the lives of many citizens who otherwise would have little or no contact with institutions of higher learning. Middle-class schools also provide the nation's so-called "2+2" programs. These allow students to take the first two years of courses at a community college, then transfer to the university to complete the degree. Most 2+2 students, for economic or personal reasons, would not be able to earn a college degree in any other way.

Overall, middle-class universities have more impact on the day-to-day lives of average citizens than do the top-50 schools. For that reason, it is important to consider the Internet2 and NGI initiatives from the perspective of this constituency.

Importance of Internet2 and NGI to Middle-Class Universities

The written testimony of other witnesses outlines the Internet2 and NGI initiatives. A new networking infrastructure is essential to the future growth of our nation. Quite simply, today's one-size-fits-all Internet doesn't meet national needs. Network congestion and outages are increasing, largely due to "growing pains" as technology that was developed for modest numbers of academic users is extended to the increasingly popular, commodity Internet.

The degradation in Internet service has become critical. It's not fair that personal entertainment traffic should slow down important research projects, or handicap the transmission of educational videos or course materials. At the same time, it's not fair that individuals seeking personal entertainment should have to bear the expense of high-speed research and education connections. As a nation, we need to develop networking technology that is capable of responding to different levels of need.

In particular, I'd like to draw your attention to four areas where Internet2/NGI will have major impact on middle- class universities, potentially changing our whole landscape for research and education.

Impact #1: Improved access to national research infrastructure. One reason that middle-class institutions don't get as many research dollars is that we don't have a strong existing infrastructure of specialized research laboratories and equipment. A network that offered high-speed access to facilities at top-50 universities and federal sites, however, would make the issue of location irrelevant, allowing us to compete for research funding on a more equal footing. The same type of network would also let us carry out long-distance collaborations with researchers at other sites - and we do have something to give to those collaborations, as well as something to get.

Impact #2: Ability to develop and deliver more flexible educational programs. Already, middle-class universities account for most of the country's innovative degree programs. Some 2+2 programs, such as the one at Oregon State, even make it possible to complete the degree through video- or other distance-learning media. But to extend access to more citizens, in more flexible ways, we must be able to deliver video or still images to a hundred locations at once, without bringing the network to its knees! This simply won't be possible without increased network bandwidth and new multicast protocols, both key components of Internet2/NGI.

Impact #3: Broader access to specialized information and services. Extension will continue to be a key endeavor for middle-class schools. For example, my university deploys over 200 extension field faculty, covering all 36 counties in the state. Each year, they make over 600,000 fact-to-face contacts with rural Oregonians, providing a wide range of specialized information and training. The 30,000 volunteers statewide who assist extension faculty donate an additional 1.2 million hours per year. Right now, an individual must go to the county extension office and request information from an agent, who then accesses it using the communications links which we have installed - at a very significant cost - in every office.

Most of us would agree that this is a sub-optimal solution. Citizens need to be able to access extension services from home (if they have a computer and modem) or from public-access terminals in convenient locations (such as community schools or libraries). It's important to note that the very persons who most need extension are those who live in remote areas, where telecommunications are most limited and most costly. Such access will be out of the question until network technology is capable of using communications lines more cost-effectively.

Impact #4: Ability to extend the scope of lifelong learning activities. Innovative educational programs are also being held back by network inadequacies. For example, Oregon State recently created an Alumni College to give former students lifelong access to courses that are delivered at locations around the state. Right now, both teachers and students must travel to course sites. If we had expanded network capabilities and technologies allowing teachers and students to interact effectively over the network, we could individualize learning opportunities without requiring physical contact.

Federal Funding Is Essential

The nation does not just need a faster version of today's Internet. Rather, we need hardware and software technologies that are intended from the outset to support millions of users at potentially millions of locations dispersed across the country. Those were never the goals of the first Internet, and existing technology simply cannot meet what has grown to be a national-scale need. There are other limitations, as well, which cannot be overcome without the new R&D initiatives proposed by Internet2 and NGI.

Current technology has several glaring deficiencies that adversely affect the kinds of programs I have described, such as extending educational services into remote areas. The mechanisms just aren't available to broadcast a single message or stream of messages to many sites in any efficient way, nor can existing technology support differential transmission speeds to meet different types of use. Without such capabilities, it isn't feasible to distribute multimedia courseware, interactive discussions with instructors or specialists, or any of the exciting new information dissemination techniques we all want to make available for isolated citizens. Thus, both infrastructure and R&D are necessary if we are to move networking to a higher level.

Internet2 and NGI cannot succeed without the injection of federal funding. Although Internet2 has managed to raise a significant sum from its academic and industry partners, we must recognize that university budgets - particularly at public institutions - are already stretched to the maximum. Even the universities that are best endowed financially cannot afford to foot the bill for on-campus networking infrastructure, plus a national network backbone and connection to it, plus the basic research and development efforts that will be required to develop the next generation of national networks.

I would add that it is not in our national interests to expect industry to bear the costs of new networking technologies. Only a few telecommunications giants are in the position of being able to invest such large sums in basic R&D efforts. There are two reasons why this would be unsatisfactory. First, there is no economic incentive for a company investing hundreds of millions of dollars to ensure that its equipment and technology are compatible with those of other companies. Yet interoperability is absolutely essential if we are to use the future networks for dispensing education, public services, or even raw information to citizens in all areas of the country. Second, by raising the entry fees for networking so high, we effectively exclude small companies or startups from participating. It was the use of public R&D funding for the original Internet that made it possible for so many new and competitive companies to emerge in the U.S. networking market.

Issues of Concern to Middle-Class Universities

The potential influence of Internet2/NGI is truly exciting, and will ultimately have impact on all sectors of the nation, not just research and education. However, there are also some issues that are of serious concern to middle-class universities.

Concern #1: Who will ensure that the new technology is really usable? We all know that there are risk factors in any new technology. My research area is usability engineering - that is, how to transition high-performance computing and network communications effectively into the lives of people who aren't computer scientists. You don't meet the needs of the masses by deploying early technology prematurely; that only drives users away. It's essential to "wrap" new technology in a cloak of usability. To do this, we must first stabilize the technology and understand the likely barriers to its acceptance; only then can we craft appropriate learning and deployment strategies. Unfortunately, the Internet2 and NGI plans have focused exclusively on developing, testing, and demonstrating the needed technology. There is no evidence that anyone is seriously considering the sociological aspects in their planning processes.

Concern #2: Who will ensure that the participation of universities and federal sites is balanced? The NGI mission statement refers to "100 universities and national laboratories." Most people interpret that as meaning "mostly universities, plus a few national labs," but the actual agency plans outlined in the NGI draft proposal almost reverse those proportions, making it "mostly federal sites, plus a few universities." Such a strategy would be seriously off-balance. Historically, the main networking technology developments have occurred in the academic environment (the federal labs have indeed contributed, but mostly in the role of "real-world" stress-testing of technology emerging from academia). Further, university R&D has been more cost-effective than that carried out by industry or federal labs, and the results of university R&D typically roll over more quickly into the private and public sectors. Universities, then, should play a more pivotal role than the NGI plan currently lays out.

Concern #3: Who will ensure that all major stakeholders are included? Even if a larger number of universities can participate, it won't be possible to locate gigapops at every university (or even in every state). This is likely to be a problem. Only the upper-class institutions will have significant resources to bring to the table. Does that mean they will be the only key players? It is important that there be a concrete plan for including universities that are more representative of national needs, such as universities that serve large rural constituencies, universities that themselves aren't located in major telecommunications cities, and universities that focus on non-traditional education (e.g., 2+2 and extension). As the Internet2 gigapop architecture is currently outlined, isolated - in the sense of telecommunications - institutions are being asked to band together in proposing gigapops. Since the cost of linking Oregon State and Montana State, for example, is well beyond their institutional capabilities, that strategy will have the effect of disenfranchising middle-class institutions unless the major telecommunications carriers can be brought to the table.

The NGI situation is potentially even worse. As currently proposed, each agency will decide independently which, if any, universities are connected to their high-speed network infrastructures (which now connect federal research installations). Unless a specific plan is put in place for such connections, access by universities to federal research facilities may not improve at all. Further, the fact that traditionally it is the upper-class universities that have served as the largest research sites for NASA, DARPA, and DOE, it would appear likely that these institutions may end up with multiple connections to new networks, while middle-class schools remain without access.

Concern #4: Who will ensure that the NGI's inter-agency collaboration really works? It's always difficult to manage collaborations that span disciplines and organizations. It's particularly hard when the parties are competing against each other for funding (as I know firsthand from leading industry-wide consortia and standards efforts). The situation calls for what might be termed "benevolent oversight," to ensure that parties don't duplicate or undermine each other's efforts. The NGI plan, however, identifies all participating agencies as partners and does not establish any clear line of authority. While the initiative would still contribute important advances if each agency were to simply carry out its own piece of the work, the real potential can only be realized if those pieces fit together well. That requires oversight and authority, as well as coordination.

Conclusion

Mr. Chairman, I conclude with a brief summary. The Internet2 and NGI initiatives are essential to our nation's future. Ambitious R&D in networking is the only way we can retain our competitive edge. A significant extension of today's networking capabilities will also make it possible to re-shape our national educational and research landscapes. Most importantly, aggressive deployment of new networking technology is the key to providing universal access, especially for extending educational, government, and health services to remote areas.

There is no doubt that Internet2/NGI will significantly advance the state of networking in the United States. Ultimately, it will affect the lives of all Americans. Its fullest potential cannot be realized, however, unless some key issues are addressed: ensuring usability to accelerate the transitioning of new technology to a broad audience, bringing the participation of academic institutions more in balance with the stated objectives, ensuring that all major stakeholders are represented, and assigning responsibility for cross-agency oversight.

Dr. Pancake's responses to follow-up questions posed by Senator Conrad Burns, chairman of the subcommittee

Article appearing in SDSCwire, Vol. 3, Issue 12, about the Senate testimony

Testimony of other panelists at the Senate hearing on Internet2 and the Next Generation Internet

Internet2: Basic Facts and Terminology (including links to other relevant Web pages)

For a brief history of the Internet, see http://www.isoc.org/isoc/publications/oti/articles/seeding.html