Current | 2005 | 2004 | 2003

3 December 2003

Add Another Zero

An Interview with Larry Smarr of CENIC, California

Larry Smarr is a pioneer in building national information
infrastructure to support academic research, governmental functions,
and industrial competitiveness. In 1985 Dr. Smarr became the founding
Director of the National Center for Supercomputing Applications
(NCSA) at the University of Illinois at Urbana-Champaign (UIUC). In
1997 he took on additional responsibility as founding Director of the
National Computational Science Alliance, composed of over fifty
colleges and universities, government labs, and corporations linked
with NCSA in a national-scale virtual enterprise to prototype the
information infrastructure of the twenty-first century. In 2000 Dr.
Smarr moved to the Department of Computer Science and Engineering in
the University of California at San Diego (UCSD) Jacobs School of
Engineering, where he holds the Harry E. Gruber Chair. Shortly after
joining the UCSD faculty, Dr. Smarr was named the founding Director
of the California Institute for Telecommunications and Information
Technology-known as Cal-(IT)2-which spans the Universities of
California at San Diego and Irvine.

Dr. Smarr received his Ph.D. from the University of Texas at Austin
and conducted observational, theoretical, and computational
astrophysical sciences research for fifteen years before becoming
Director of NCSA. He is a member of the National Academy of
Engineering and a Fellow of the American Physical Society and the
American Academy of Arts and Sciences. In 1990 he received the
Franklin Institute's Delmer S. Fahrney Gold Medal for Leadership in
Science or Technology. Dr. Smarr is currently a member of the NASA
Advisory Council, Chair of the NASA Earth System Science and
Applications Advisory Committee, and a member of the Advisory
Committee to the Director, NIH. He also served on the President's
Information Technology Advisory Committee (PITAC).

In the following conversation with EDUCAUSE Review,Dr. Smarr
discusses the future of broadband, the "Gigabit or Bust" initiative
in California, and Grid technologies, among other topics.

EDUCAUSE Review: Larry, your career has covered a wide range of
interests and disciplines. How would you characterize your
professional goals and interests at this stage of your career?

Smarr: I've been privileged to work with many pioneers not only in
computer science and my original fields of physics and astrophysics,
but more broadly in earth sciences, environment, medicine, biology,
and nanotechnology. I guess I think of myself as a perpetual student
because there is so much to learn and yet even more is being
discovered.

My current position is Director of the California Institute for
Telecommunications and Information Technology. In a way, this is a
dream position. At Cal-(IT)2, we were able to start with a blank
piece of paper and create both an institutional structure and a
disciplinary set of topics that we're going to be examining over the
next five, ten, twenty years. We picked the fundamental goal of
looking at the evolution of the Internet, enabling us to explore a
lot of wireless technology areas as the Internet moves throughout the
physical world. Many of the devices that the wireless Internet will
connect will be based on micro-electrical-mechanical systems (MEMS)
technologies and ultimately nano-based technologies. That is why, in
the two Cal-(IT)2 buildings at the University of California at San
Diego and the University of California at Irvine, there are
clean-room facilities to support bio-MEMS, material characterization,
nano-fabrication, and so forth. This has allowed me personally to
branch out into areas of physics and chemistry and
electrical-engineering optics that I hadn't studied before.

On the other hand, we're applying these technologies to areas that I
think will evolve a great deal as the Internet evolves: for example,
we will be able to put massive sensor nets into the environment, our
transportation systems will be based on peer-to-peer sharing of
information among the automobiles and the infrastructure of the
highways, and even our own bodies will move online, with biologically
appropriate sensors that can measure our vital signs and report over
the Internet in a secure and private fashion. Another emerging arena
is the world of networked computer gaming, in which hundreds of
thousands of people are choosing to live their lives in a virtual
cyber-world and to create a cyber-civilization in which they have
totally different roles and personas than they do in the physical
realm.

I guess if I had to pick a phrase that characterizes my future
interests, it would be this intersection of the physical world and
the cyber-world, this blend of atoms and bits that is beyond anything
we've yet experienced as humans. It is, I think, a very deep area and
one that will involve not just technical folks, but also performance
artists, social scientists, even philosophers.

EDUCAUSE Review: Higher education and government will command a
smaller and smaller portion of U.S. networking infrastructure as
broadband moves to the home. Similarly, U. S. networks will represent
a decreasing portion of global infrastructure as other countries
build out. Are there important implications here for leadership and
innovation?

Smarr: Clearly, the ability to get true broadband to hundreds of
millions of homes and small businesses in the United States and
throughout the world will be the next big driver of the economy.
We've seen studies over the last five years from industry groups,
from the National Research Council on the academic side, and from
states that are developing projects like California's "Gigabit or
Bust" initiative. All agree: the current wired Internet coupled to
the personal computer is an S-curve transition that is on its flat
mature top. We therefore cannot expect to see the high rates of
growth that derived from the coevolution of the wired Internet and
the personal computer, a development that drove the economic miracle
and bull market of 1982 through 2000.

We have to look at new S-curves. One of them I described before is
the wireless Internet, which links to cell phones, PDAs, cars, and
sensor nets. Another new S-curve is broadband, either wired or
wireless, to the home. This represents a new coevolution of
technologies that many people believe will pull the cork out of the
bottle and release the next wave of economic growth. One of the big
barriers is caused by regulations, and therefore change is much
slower than it should be. Our economy suffers as a result.

Already a number of countries, like Korea, are very far ahead of the
United States in the penetration of broadband to the home-and of
course by "broadband to the home" today we typically mean only one
megabit per second or less. Yet many states in the United States are
now beginning to experiment with different approaches to accelerating
the broadband buildout. And that's the great thing about the United
States: we have fifty laboratories for innovation. States like
California, with the "Gigabit or Bust" program, are ambitiously
attacking the problem. So I think we're going to have winners and
losers. I think the countries and the states that are aggressive
about getting broadband to the home-creating a new market with
literally tens of millions, maybe eventually one hundred million
participants, and creating a whole new set of devices in the home,
not just entertainment but Internet-linked washing machines, dryers,
refrigerators-are going to unleash a great wave of new companies and
all kinds of new ideas. So we're pushing very hard for the idea at
Cal-(IT)2 and certainly here in California.

EDUCAUSE Review: The goal of the "Gigabit or Bust" initiative is to
bring one gigabit of broadband per second to every home, classroom,
and business in California by 2010. How realistic a goal is that?

Smarr: I have to take some blame for the "Gigabit or Bust" slogan.
When the Corporation for Education Network Initiatives in California
(CENIC) and the Next Generation Internet (NGI) Roundtable were
considering broadband to the home, we looked at all the previous
reports. Most reports talked about a national decadal goal of 100
million bits per second going to the home, which would be an
improvement of roughly one hundred times the speed of today's
broadband. So that would not be an insignificant change. But I felt
that if California is going to think of itself as a leader, then it
needed to "add another zero." Besides, a gigabit seems like such a
nice, round number, and "one hundred megabits" just doesn't roll off
the tongue as easily.

But seriously, think how absurd the situation is today. If you buy a
Macintosh laptop today, it comes with a built-in gigabit Ethernet,
included in the price. So, our personal computers in 2003 have a
gigabit input or output, and yet people are saying that in seven
years we can't get that kind of bandwidth to our homes? In seven
years, what do you think laptops are going to have for bandwidth?
There's this crazy mismatch right now between the "last-mile" problem
(which isn't a mile at all-it's more like the
last-twenty-or-thirty-feet problem, from the curb to the house), and
as a result we have these islands of data in our computers, in our
servers, and we have this absurd bandwidth bottleneck between them.
We have to smash that bottleneck and unleash the enormous
peer-to-peer bandwidth capability that the intrinsic system of
servers and personal computers allows for.

The "Gigabit or Bust" initiative is ambitious, and it is
controversial. So one of the things that the NGI Roundtable and CENIC
have come up with-and that I'm very excited about-is an annual
competition in California, in a variety of categories, for the best
examples of communities, of school districts, and of companies that
are leading the charge toward the "Gigabit or Bust" goal. We're
publicizing these awards because in these technology transitions, the
biggest barrier is that there is no model to follow. A few good
examples will inspire others. Thus this contest that CENIC and the
NGI Roundtable have established could be one of the most important
reasons that California gets, at the least, very close to the goal.
In 2010 there will be many houses with a gigabit per second, and
there will be some houses with only up to 100 million bits per
second-but if we can get to that point, it'll be a great success.

EDUCAUSE Review: High-end research is certainly one driver of
ubiquitous broadband, but do you believe that pressures from the
popular culture-specifically the Net generation-may be equally
important in building that future infrastructure?

Smarr: Almost every component of today's information infrastructure
started as a result of federal funding of university research. I
believe we will get a lot of technology momentum driving for gigabit
to the home from experiments on college campuses, which are, after
all, small towns. At UCSD, 40,000 people are associated with the
campus. There are streets and utilities, and we have to dig from the
curb to the dorms, but we already have 100 megabits to lots of
students' rooms at UCSD in 2003, roughly 100 times the bandwidth of
what passes for "broadband" to the home today. College campuses
therefore give us a time machine because they offer us a "living
laboratory" of how our society may live five to ten years from now.
People who are interested in this issue of broadband to the home
should be studying our college campuses intensively today, yet I
don't think there's nearly enough of that being done.

On the other hand, look at the way that the kids are driving the
network gaming industry: requirements for computer graphics to
support computer gaming are now exceeding those of feature-length
movies, and the revenue streams are beginning to move in that
direction. There is a wonderfully insidious, driving force in every
home with children in California or anywhere else in the country or
the world. A kid without broadband at home can't get to the
beautiful, high-resolution, 3-D graphical worlds of these modern
networked games. This kid is going to say: "I can't have a proper
social life-I can't grow up as a good kid with my peers-unless I have
high bandwidth at home." That kind of grass-roots, bottom-up pressure
will be one of the most important drivers.

EDUCAUSE Review: A recent "Gigabit or Bust" roundtable focused on
several public-policy areas that may offer potential "killer apps"
for ubiquitous broadband. Do you have any favorite killer apps in
health, government, business, and the like?

Smarr: Again, I think we can study the sociology of how a broad
citizen base will interact with a ubiquitous broadband environment by
looking at college campuses today. Internet2 connects about 200
campuses, with between 5,000 and 50,000 or so students at each of
these campuses. They have 10 or 100 megabits to their rooms, and
their personal computers typically have 50- to 100-gigabyte drives
and 2-3-gigahertz pentium chips. Then the campuses are all hooked to
each other over Internet2, which is by now gigabit to tens of gigabit
connectivity. So if you want to ask about the broadband killer app,
you should look at this broadband "living laboratory" connected over
vastly different geographic distances and interests. What you see is
that in this laboratory, the killer app that has emerged is the
sharing of multimedia objects. And this, of course, has raised very
challenging intellectual property issues. But the fact that we have a
controversy over intellectual property shouldn't blind us to the fact
that if people are given broadband access, there will be an intense
hunger for the sharing of multimedia objects such as music and video.
This is such an overwhelming drive that people are willing to go to
jail for it. Now that's what I call a true killer app.

EDUCAUSE Review: You coined the term "metacomputing" in 1988 to
describe the integrated utilization of a variety of distributed
computing resources. How important are such things as the Grid
community, the Grid toolbox, and the Grid infrastructure in 2003 and
beyond?

Smarr: The transition in the Web from the early 1990s to the late
1990s is a classic S curve. A similar transition is happening for the
Grid between the mid-1990s and the end of this decade. The early
days, at the bottom of the letter S, form the era of early adopters.
There is typically a lot of experimentation: different people have
different ideas about the right way to do something complex-for
example, middleware. Next is a period of consolidation, during which
those ideas get sorted out and people decide that one particular
brand of software, or some standard, is the way to go. After that is
take-off-the knee of the S-curve at the bottom. I think we're about
at that point now with Grid middleware. This middleware is the
"operating system" of the metacomputer I envisioned almost fifteen
years ago-but with the ability to authenticate and, in a secure
fashion, to reserve computing storage, networking, visualization,
files, data sets, people, and instruments. This middleware enables
the integration, from everything that's tied together by the
Internet, of a specific electronic metacomputer that you and the
other people hooked to it need to use for the next ten minutes, or
one second, or two hours.

I think it's clear that Globus, which was developed by Ian Foster and
Carl Kesselman, has emerged as the gold standard in Grid middleware.
It is being adopted by the Europeans, by the United Kingdom, and by
many Grid efforts in Asia and in the United States as well. One of my
concerns is that the United States is lagging in the race to build
out Grid middleware under its large shared science projects. But the
National Science Foundation (NSF) will soon launch a distributed
cyber-infrastructure initiative, which should help us catch up to the
Europeans and the British. It's always healthy to have some
competition, and you don't necessarily always want to be the first to
adopt something. But many of these ideas were developed in the United
States, and I think it would be a shame if the United States did not
reap the benefits. On the other hand, it's very exciting to see the
internationalization so early in the history of the Grid. So really
the only issue is whether the United States will be able to carry its
weight as the global scientific Grid develops as the essential
information infrastructure foundation for discipline after
discipline-from particle physics to earthquake engineering to ecology
to astronomy. I'm very encouraged by what I see happening at the NSF
right now. There has also been pioneering work at the Department of
Energy and also very good work coming from the National Institutes of
Health and from NASA.

EDUCAUSE Review: Will the term "supercomputer center" become obsolete
as Grid technologies support the widespread distribution of
computational resources, support staff, and users? Will the focus
shift to applications in "virtual centers of collaboration" as
computing power becomes a utility? Where will all of this lead?

Smarr: I find it difficult to predict the future without
understanding the past. When you look at previous infrastructures-the
railroads, the electrical power system, the air traffic system-what
you find is that as the infrastructure develops, it is "lumpy," with
very different sizes of "lumps." I think the easiest way to
understand this is to look at the air transport system. There are
many airports around the world. I used to fly from Willard Airport, a
very small airport in Urbana-Champaign, Illinois, up to Chicago's
O'Hare, one of the largest and busiest airports in the world. Just
because there's a distributed set of airports doesn't mean that
they're the same size. Any developed, mature infrastructure will
display this inhomogeneous distribution of capability.

Consider your nervous system. At the top of your spine is this big
neural "lump," called the brain, which has specialized sections:
cerebellum, cerebrum, and so forth. You have a medium-sized neural
lump called the solar plexus; you have different sizes of nodes of
neural-processing capacity all through your body, down to the
individual sense cells in your fingers that feel pain or cold or hot.

Likewise, it's completely clear to me that this distributed computing
capability is beginning to develop in the Grid: from individual
personal computers to Linux clusters in our laboratories, to the
super-nodes on the Grid, which contain the most extreme computing or
storage capability or the most extreme instruments.

In a Grid world, we are going to need a few large national centers
every bit as much as we do today. In fact, I would argue that we will
need them even more because someone has to help design, prototype,
and build out these large-scale Grid engineering systems. Examples of
these hyper-data-intensive shared scientific systems would include

• CERN's Large Hadron Collider particle accelerator, with the
  hundreds of universities and tens of thousands of users that share it,
  
• NASA's current fleet of twenty Earth-observing satellites sending
  down a terabyte per day of new data that goes out through NASA's
  Earth Observing System, delivering millions of data products a year
  to hundreds of thousands of end users, and
  
• the National Institutes of Health's Biomedical Imaging Research
  Network, which connects many universities forming a federated
  repository of multi-scale brain images.

All these and many other shared scientific systems ultimately should
depend on a universal standards-based, distributed
cyber-infrastructure that globally ties together all of these sources
of data with all of the consumers of data and with all of our
scientific instruments.

To build such a Grid, we're going to need large centers like the
supercomputer centers again. The most important thing about the
supercomputer centers is not that they have supercomputers, though
that certainly is important. It is that they bring together a large
group of experienced and dedicated technical professionals, from many
technical specialties, who create a team that serves the U.S. user
community. These teams have built a great deal of what we now know as
our information infrastructure. We're not going to need fewer of
these people because we're building a more complex system-we're going
to need more of them.

Does this mean that we'll need only a few large centers? No. There
will be a logarithmic-distribution function of capability spread
across the campuses. In addition to the larger centers, there will be
centers of specific focus on many campuses. There will be many Linux
clusters, with associated storage and visualization in university
laboratories studying chemistry, physics, biology, astronomy, and the
engineering field sciences. There will be a large, distributed set of
different-scale centers building out the cyberinfrastructure. But I
think there are going to have to be some premier large centers that
do the national work because that's what I see in every
infrastructure that I look at.

EDUCAUSE Review: Your current responsibilities at Cal-(IT)2 involve
building collaborative research agendas and working relationships
among a wide range of accomplished scholars. The sociological
challenges in all of this must be interesting. In managing this sort
of research, are you finding some patterns or frameworks that others
might adopt?

Smarr: Governor Gray Davis declared, in December 2000, that
California was taking the bold initiative of establishing four
institutes for science and innovation: Cal-(IT)2; the California
NanoSystems Institute (CNSI); the Center for Information Technology
Research in the Interest of Society (CITRIS); and the Institute for
Bioengineering, Biotechnology, and Quantitative Biomedical Research
(QB3). Richard C. Atkinson, president of the UC System, endorsed the
idea, and a bipartisan vote in 2002 by the California legislature
approved the capital funds for these new institutes. We also received
a lot of support from industry and from faculty winning federal
government grants. I had learned a lot from being involved in
previous national experiments of this sort, so I looked at these
earlier experiences and tried to understand what would be a natural
way to organize the two hundred faculty involved in Cal-(IT)2 across
not just the San Diego and Irvine campuses but also the San Diego
State University, University of Southern California, and other
campuses that are working on particular projects around the country.

One of the big differences between NCSA and Cal-(IT)2 is that in
NCSA, most of the members were academic professionals, with only a
few faculty. Cal-(IT)2, in contrast, is a faculty-driven organization
with a few technical professionals. In fact, Cal-(IT)2 needs many
more technical professionals, but the operating budget won't allow
for it right now. So I knew from the start that we would have a
number of levels, all of which would be run by faculty. We developed
a "layer concept." At the bottom are the new materials and devices
that the future Internet will require; next is the networked
infrastructure; and then come the interfaces and software systems.
Then we have our four driving applications areas: environment and
civil infrastructure; intelligent transportation and telematics;
digitally enabled genomic medicine; and new media arts. We then added
policy, management, and socioeconomic issues on top. Education and
industry cut through all these layers. Each layer or applications
area is led by two faculty researchers, one from each primary campus.
All "layer leaders" report to the two campus Division Directors, who
in turn report to me. Working with me are Ron Graham, Chief
Scientist, and Stephanie Sides, Director of Communication. This has
worked pretty well. With roughly ten faculty for each layer leader
and with about ten layer leaders for each division director, there is
a manageable numerical scale. And yet the structure allows for great
decentralization of innovation. My job, meanwhile, is to try to see
the emerging themes that come from such a broad, cross-disciplinary
set of scientists. I spend time educating myself so that I can
understand the trends in many of the underlying technologies or
science areas, understand where the federal agencies are going,
understand what the global competition is like, and then try to lead
us in directions that are productive.

The Cal-(IT)2 Layered Collaborative Structure
The Cal-(IT)2 Layer Collaborative Structure

I think creating this collaborative structure is fairly natural. I
can imagine it being applied to any campus. I think the scale is
perhaps bigger than most campuses would want to undertake, but it
could be used for essentially any topic.

EDUCAUSE Review: Some applied disciplines find themselves stuck in
the pre-digital environment, in terms of both equipment and
processes. Based on your work with researchers in the medical
community, what role do you see for higher education in testing and
promoting digitally based "sensor nets" and real-time feedback,
diagnosis, and treatment?

Smarr: Our country underestimates the value of campuses as places to
work out innovation in "systems." We've optimized our campus research
environment for individual professors to do curiosity-driven
research, and that certainly will continue to be the foundation on
which advances are built. But somehow, along the way, we've lost the
concept of "systems" as being an important area for research. When we
look back at what John Hennessy was doing at Stanford when he was
head of the MIPS project in the 1980s, we see that campuses can be
engaged in very complex systems research. I think we're going to see
more and more of this kind of systems research join the
hyper-specialized individual research, but we're also going to need
new institutions to support it. This is one of the reasons that
Cal-(IT)2 is an interesting experiment-to see if a sustained
collaborative framework can be built into a preexisting campus
structure of deans and departments and individual researchers.

For example, Bill Griswold, in the Department of Computer Science &
Engineering at UCSD, has handed out nearly one thousand Windows CE
Pocket PCs with Wi-Fi and spatial software to students. Thus a small
community of students, all undergraduates, are living in the kind of
world that will exist in the future-not that far in the future,
within the next year or two-when all of our cell phones will be
geo-located. Many already are. Then, when you are working with people
over the Internet, you will have not only their names but also their
locations, if they want to disclose that information to you.

Where else can we do this kind of experiment on such a scale? And the
same is true in medicine. I think many medical schools will realize
that they can begin to prototype wireless embedded medical sensors,
digital medical records, and the ability to do data-mining across
large numbers of human medical records-things that we're not capable
of doing today. I would like to see much more emphasis on this kind
of research on college campuses.

EDUCAUSE Review: Many institutions have IT strategic plans that
involve, among other things, wireless networks. Do you think that
Wi-Fi is the next big thing?

Smarr: I certainly think that mobile Internet is the next big thing.
Whether the specific unlicensed spectrum, 802.11 and its variants, is
the next big thing-whether it's cellular Internet like 1xRTT in the
United States or whether it's a bridging of these-that's for the
market to sort out. But there is an explosive growth in mobile
connectivity. As I recall, a recent study found that across public
universities in 2002, roughly 20 percent of the educational area of
classrooms and dorms and eating places was covered with Wi-Fi. Of
course, on some campuses the percentage is much higher. Three years
ago Carnegie Mellon was already deploying a great deal of Wi-Fi. At
UCSD, I think 96 percent of the educational areas are covered by
Wi-Fi. So there are pioneering schools that are learning how to live
in a world in which the Internet is everywhere. And people don't
really understand how big a change that is. It means that you can put
a small device anyplace, and the device will be able to connect to
the Internet because the Internet is already there in wireless form.
On the other hand, to get bits from one place to another using
wireless as an underlying medium, rather than electrons going down
copper wire or photons going down clear glass, is vastly more
complex. We have a great deal yet to learn about the electrical
engineering of how to get the Internet throughout the physical world.

EDUCAUSE Review: You are very interested in nanotechnologies and in
the general ideas of embedded intelligence and communications. When
intelligent communicating devices get really small, what will change
in our living, working, and learning environments and processes?

Smarr: One of the biggest mistakes people make when they predict the
future of the Internet is that they consider only the networking
aspect. But I think what's going to happen over the next two decades
is that we're going to see an accelerating rate of change, similar to
what Ray Kurzweil has talked about. This is going to induce what I
call "the perfect storm."

In the movie The Perfect Storm, a boat got into trouble because
several storms moved together and merged into a super-storm, which
created chaos and violence on a scale that no one expected. I propose
that something similar is going to happen with the Internet. The
storms are (1) information technology and telecommunication, (2)
post-genomic biology, and (3) nanotechnology, coming from
engineering, physics, and chemistry. Each is a huge revolution in its
own right. But they're all happening at the same time, and they're
all going to merge into one large storm of info-bio-nano-technology.

The world that we have built in academia, a world based on
specialization, will have a real problem dealing with this perfect
storm. Doctors don't think they have to know anything about
information technology or physics. Physicists don't think they need
to know anything about biology or information technology. And
computer scientists don't think they need to know anything about
biology or physics. By contrast, kids understand that this perfect
storm is brewing; they are sliding across these stovepipes and are
picking up biology, physics, chemistry, and information technology.
They are living on the Net, sharing all these advances, and they're
going to be much better prepared for this perfect storm. They're
going to ride it out. But I'm afraid that many people-specialized
faculty, in particular-are going to be in for a rude shock.

EDUCAUSE Review: What are some of the policy, fiscal, or technical
challenges that data-intensive sciences will pose for colleges and
universities in the years ahead?

Smarr: As the data Grid develops both nationally and internationally,
campuses are going to have to adapt to it-much as they adapted in the
mid-1980s when the NSF decided to link the five NSF supercomputer
centers together using TCP/IP derived from the ARPANet (Advanced
Research Projects Agency Network). Campuses had to decide whether or
not they wanted to participate in providing access for their faculty
to these remote resources. It was quite amazing to watch the reaction
because there were actually only a handful of potential supercomputer
users on each campus-normally not enough for institutions to get
bulldozers to come out to trench the quads and for them to make
capital investments in laying cable. But they did. And campuses did
so because they knew they would lose their best faculty if they did
not. And so, without anybody having to tell institutional leaders
that they needed to carry out some infrastructure changes, they
naturally did it.

I think the same thing is going to happen as we move from this
supercomputer-driven to a data-intensive world. There are probably
ten experimentalists or observers among the scientific community for
every one theorist. As we enter this data-intensive world, we're
going to get a much larger participation than we did fifteen years
ago in the supercomputer world, which was largely driven by
theorists. What we will need very soon is a standardized laboratory
environment-integrating computing, storage, and visualization-that is
very easy to set up.

Because the Grid is data-driven, there will be a new utility needed,
a data-storage utility, which is going to have to be provided by the
campus as a whole. If you've got all of these faculty on campus
pulling down large data sets from remote federal repositories through
the campus gateway to their individual laboratories, and you don't
have a large data cache for the whole campus, then you're going to
have to spend a ton of money on a giant gateway. Because every time
faculty want to get a new data set, they're going to run out of space
in their lab, so they will erase the older data. Then when they want
to look at that older data again, they're going to run the same data
set through the gateway, and this just doesn't make any technical
sense. Campuses need large storage-say, one hundred terabytes or more
of rotating storage that is available to anyone on campus. Then
individual labs might need only a few terabytes for their own daily
use of data, with every data set being stored in the shared campus
cache for perhaps six months or so. When the users want the data
again, they need only to go back to the center of the campus and not
halfway across the country.

This doesn't sound like such a big challenge. But if two hundred
universities are trying to figure out this storage architecture-each
on its own-this is not going to be very efficient. Soon campuses will
need to begin to think about finding common solutions. One thing that
we need more of in the United States is dialogue among the chief
information officers about these standardized solutions-whether for
the individual laboratory or across campus. CIOs do not often get
involved in what scientists put in individual laboratories, but maybe
they should. If we could find more common solutions, we might save
everybody a lot of time, and we could get on with doing the science.

Steve Daigle, Senior Research Associate, Information Technology
Services, California State University Office of the Chancellor,
served as contributing writer and facilitator for this interview.