![[Internet2 Logo]](http://www.internet2.edu/Images/i2-logo.gif){kind=logo}
Bill Decker, University of Iowa, Bill_Decker@uiowa.edu
Bill Graves (Chair), University of North Carolina at Chapel
Hill, Bill_Graves@unc.edu
Ted Hanss (Staff Lead), University of Michigan, ted@umich.edu
John Kolb, Rensselaer Polytechnic Institute, kolbj@rpi.edu
Dave Lambert, Cornell University, hdl2@cornell.edu
Bill Lewis, Arizona State University, William.Lewis@asu.edu
Clifford Lynch, University of California, Clifford.Lynch@ucop.edu
Ray Neff, Case Western Reserve University, rkn@po.cwru.edu
Chris Peebles, Indiana University, peebles@indiana.edu
Ed Sharp, University of Utah, sharp@cc.utah.edu
Don Spicer, Vanderbilt University, Donald.Spicer@vanderbilt.edu
Craig Summerhill, Coalition for Networked Information, craig@cni.org
Mely Tynan, University of Arizona, tyson@arizona.edu
Dan Updegrove, Yale University, danu@yale.edu
Al Weis, Advanced Network & Services, weis@advanced.org
Tom West, California State University,
twest@calstate.edu
In a background paper released with his speech of October 10,
1996, announcing the Clinton-Gore Next Generation Internet Initiative,
President Clinton included the following key goal:
"Demonstrate new applications that meet important national
goals and missions: Higher-speed, more advanced networks will
enable a new generation of applications that support scientific
research, national security, distance education, environmental
monitoring, and health care."
A few days earlier on October 1, 1996, a sizable group of universities
had created the Internet 2 Project for the primary purpose of
enabling the development of a new generation of network applications
in support of scientific research, distance education, environmental
monitoring, health care, and digital libraries --see "Examples
of Internet 2 Applications and Application Development Tools"
below. Once again the nation's political and academic leadership
were joining forces for the common good to advance the nation's
economic and social goals.
Indeed, only a few years have passed since modest federal investments
in NSFNet provided leverage for a much larger total investment
in campus-based network infrastructure. These investments by higher
education and a few key federal, state, and corporate partners
were designed to enrich the nation's research infrastructure,
but they also quickly resulted in a range of unanticipated, broadly
useful applications in the global academic community. The result
was the first general purpose (global) Internet. Soon thereafter,
the Internet became an integrated set of inter-networking resources
and services based on open, de facto standards and offered by
an array of competing providers in a commercial environment now
exhibiting the features of a commodity market. The World Wide
Web (Web) and its attendant browsers, with their origins also
in the research and academic communities, catapulted the Internet
to its current revolutionary status both as a social and an economic
phenomenon.
Today's most popular Internet applications followed on the heels
of development and research surrounding the network technology
itself. Today's context is quite different. Network applications
increasingly capture the nation's intellectual capital as leverage
for economic development. But the provision of bandwidth and advanced
network technology is lagging the development of applications
ultimately requiring high performance network services.
MIME-compliant email and Web servers/browsers have developed in
parallel with sophisticated tools for developing stand-alone multimedia
applications. Together these developments have raised expectations
for the network delivery of audio and video streams. Synchronous
MUDD, MOO, chat, and multicast technologies and increasingly sophisticated
asynchronous tools for collaboration and workflow have raised
expectations for using the network for collaboration based on
application sharing and desktop video teleconferencing or other
realistic real-time communication technologies. Medical applications
of these technologies are often envisioned in the popular press.
For example, the ability to deliver monitoring data with quality
of service guarantees along with high-resolution images in a secure
fashion over long distances can bring remote medical experts directly
into interactive patient care. The results of queries to online
repositories can be delivered almost instantaneously to the physician
needing to compare images while making a diagnosis. Indeed, today's
expectations include access to more general, large distributed
databases and to distant network attached instruments with provision
for the distributed analysis of their data streams, even in interactive
mode. By reducing the barriers of limited processing power and
bandwidth, analysis currently performed "off-line" could
now be done interactively with Internet 2. Geographic information
system researchers, for example, could interactively correlate
data from distributed social and physical sciences repositories.
These advanced services are applicable even for the analysis of
texts. For example, researchers could perform interactively iterative
relevance analysis with digital library content searches involving
large repositories distributed over multiple sites.
The promise of these emerging applications for distributed instruction,
collaborative research, and strikingly new forms of publication
and dissemination is compelling. In the potential of these applications,
the universities participating in the Internet 2 Project have
glimpsed the future of higher education and are determined to
seize that future for the common good of all of education. Their
goals are consistent with, and supportive of, those of the Clinton-Gore
Next Generation Internet Initiative.
Internet 2 institutions have committed to making substantial investments
in institutional and inter-institutional infrastructure in order
to develop and demonstrate compelling applications of next-generation
network technology to instruction, research, and public service.
These same institutions realize, however, that the promise of
such investments will not be fully realized until the same advanced
network services characterizing Internet 2 are extended to all
of higher education, the public schools, the workplace and, especially,
the home. Only then can the limiting boundaries of classroom,
library, and laboratory be neutralized to deliver, for example,
learner-centric distributed instruction -- the virtual curriculum.
This is one of the key reasons the Internet 2 Project is committed
to two-way technology transfer between participating institutions
and the many commercial and non-profit organizations working to
influence the future of the Internet.
Internet 2 technical specifications call for network services
incorporating the demand-driven growth of bandwidth with bandwidth
reservation services, quality of service guarantees, and advanced
forms of functionality (e.g., integration of voice, video, telemetry,
and data services). A specific design point of Internet 2, moreover,
is a provision for dynamically expanding capacity and functionality
to meet demand into the future. Can Internet 3 be far away in
Internet time? Indeed, the network and its services must be designed
not to impede or constrain applications design. The network must
become responsive to the demands of new applications, even applications
previously unimagined or currently existing only in very specific
forms (or through special accommodations). The brief history of
the Internet is replete with serendipity, which should be a design
point for Internet 2. All applications should be equally "createable."
A large part of this document - "Examples of Internet 2 Applications
and Application Development Tools" -- accordingly describes
in more detail some examples of possible applications that have
prompted universities to invest in the Internet 2 initiative.
These applications in whole or in part cannot be supported by
today's Internet connections between participating institutions
and often also require institutional intranet services not widely
available. Many of these examples will further require workstation
and operating system functionality not readily available. An applications
architecture therefore must be evolved, and companion services
and development strategies and methodologies must be promoted
to accommodate the demanding applications fueling the Internet
2 initiative.
A key strategy is to focus on applications which either require
or would be substantially enhanced by the advanced services envisioned
in conversation with the Internet 2 Engineering Working Group.
The latter include bandwidth reservation protocols and service
guarantees to mitigate latency in time-sensitive applications.
Such protocols and guarantees, for example, would enable streaming
video servers and requesters to perform beyond today's capabilities,
thereby further delivering on the promise of the Web for education,
entertainment, and business. Similarly, network-delivered learningware
incorporating streaming video clips would help deliver on the
promise of distributed instruction. A related example is the Instructional
Management System (IMS) being developed under the aegis of EDUCOMís
National Learning Infrastructure Initiative as a standard for
supporting network-delivered learningware and managing a distributed
instructional process. The IMS would be enhanced by incorporating
the synchronous application sharing and video teleconferencing
enabled by the advanced services of Internet 2. Similarly, researchers
would welcome the opportunity to remove the constraints of latency
and limited bandwidth in their experiments involving distant data
and its mathematical analysis expressed through a time-sensitive
visualization on their displays.
Among the applications more fully beyond the reach of today's
Internet are tele-immersion and a host of virtual laboratory projects.
An interesting example of a virtual laboratory project would focus
on the development of a nanomanipulator -- a natural virtual reality
interface to network-connected scanning probe microscopes, including
scanning tunneling microscopes and atomic force microscopes. Tele-immersion
would go further by allowing several participants to share a common,
realistically rendered virtual environment while communicating
in normal human fashion within that virtual environment and interacting
with a common application.
These and other examples are elaborated in more detail in the
last section of this document.
A second key strategy is to recognize that good applications are
most likely to proliferate in an environment rich in good application
development tools. Although such tools are themselves most likely
to be discovered in the context of developing specific applications,
it will be important to recognize explicitly the need the for
application development tools for Internet 2. Web servers and
browsers are examples of tools that have enabled the development
of thousands of applications of Internet technology. In the Internet
2 context, good generic tools can encourage a thousand flowers
to bloom, while it would be impossible to prioritize and develop
all the applications that researchers and educators might desire.
The aforementioned Instructional Management System is an example
of a set of protocols, middleware, and client tools that could
do for the development of media-rich distributed instruction what
Web protocols and Web client tools have done for the online publication
of information.
It will be important to identify those application development
areas which are being (or soon will be) addressed by the commercial
sector. The goal in such cases will be to provide for the participation
of Internet 2 institutions, as appropriate, in the design and
testing process for these tools and applications.
Similarly, it will be important to identify key application development
areas which are not likely to be addressed by the commercial sector
and to develop strategies for advancing these areas, including
strategies for funding and managing any attendant development
projects.
One mechanism might serve the effort well in both cases. A Request
For Partners (RFP) -- coming from a partner company, a member
institution, or some combination of member institutions and partner
companies -- could articulate the parameters of the development
and/or testing in question and establish criteria for participation.
An RFP initiated by a commercial partner should not be linked
to the sale of a product or to any overt marketing agreements.
But in all cases, RFPs should be based on the expectation of delivering
an advantage for all participants. This might be, for example,
an advantage in timely access to research or time to market for
a company and in early access to emerging technologies for Internet
2 institutions. A key premise of Internet 2 is the need for two-way
technology transfer between participating universities and their
partners representing commercial and governmental interests.
Here is a brief, abstracted example to illustrate the idea of
an RFP, this one coming from Company X.
Sponsoring Partner: Company X (division name and contact)
Application Area: Media-rich email.
Forms of assistance:
workstation(s) with digital video cameras and ….
alpha code which …
on-site technical assistance
collaboration with researchers from Company X's Research Center
Availability: Company X will select up to six Internet
2 institutions supporting IPv6 over OC-3 inter-connections for
this applications work based on informal proposals from the senior
IT officer. To be considered, campuses also must have established
minimum 25Mbps IPv6 connections to participating test sites and
must be willing to scale the tests to serve a significant client
base.
The next example represents the possibility of Internet 2 institutions
sponsoring an RFP.
Sponsoring Partners: Universities of X, Y, and Z (senior
IT officers or other contact points)
Application Area: Collaboration applications for molecular
modeling.
Available Resources:
PI and Co-PIs will submit NSF grant for additional support.
Universities have vBNS connections.
PI and Co-PIs have desktop to desktop 100 Mbps capabilities.
Needs:
Workstation systems supporting IPv6 and development tools for collaboration applications
Infrastructure electronics supporting IPv6 at 100 Mbps at the edge and OC-3 elsewhere
Other information and requirements: …
Member institutions must decide to what extent the project should
identify key application development areas and fund development
work. Regional groupings of participating institutions and their
regional partners are forming to implement the distributed GigaPoP
architecture envisioned by the Internet 2 Engineering Working
Group. These regional coalitions will surely invest in application
development. But it will also be important to identify application
development tools and application areas that should not be left
to the vagaries of regional efforts and earmark these as national
priorities deserving of investment by the Internet 2 Project and
its national partners. In the case of both national and regional
development projects, the Internet 2 Project should provide coordination
and, when Project investments are involved, management for these
development efforts. Toward this end, the project will employ
an applications staff to provide coordination and management.
A staff leader for applications development has already been hired
and will work with the Applications Working Group and participating
member institutions and partners to design and implement an applications
development architecture.
Many of the trends in programming and applications development
during the past decade will contribute significantly to the Internet
2 applications environment. Among these trends are object-oriented
programming, software componentization, object request brokering,
and dynamic run-time binding. Also significant is the trend toward
multi-tiered applications delivery with separation of data, process,
and presentation functions.
What will differentiate the Internet 2 world, however, is the
ability to migrate all of these technologies and concepts into
a fully distributed realm, away from simple and restrictive client/server
ideas, and to accelerate the rate at which the promise of these
technologies can be realized.
It too early to know what the appropriate technology or architecture
will be. But the Internet 2 Project should explore the issue of
"middleware" in a high-bandwidth, low-latency, quality-of-service
enabled network environment. Network performance should not be
degraded by wasting time in such overhead functions as directory
lookups and authentication via security servers. Does this mean
going back to hard-coding such functions into applications, and
thus raising the cost of managing applications even if they do
perform better? Or should the project revisit the assumptions
of how to implement such functions as directories and security
when deploying them for Internet 2?
A full architectural model for Internet 2 will evolve to address
the following representative concepts:
A placeholder architecture for Internet II applications might
follow the model depicted below, where the client side relies
on the component technologies mentioned in this paper for construction
while middleware services, using operating systems functionality,
communicate over the network to applications and network services.
The server-side depiction implies an n-tier model where
multiple servers may be applied to a single application. As noted
above, these architectural ideas are meant to be a starting point
for discussion, rather than a declaration of design.
Achieving a common vision will require communicating a common
set of guidelines to network providers and application developers.
The above requirements may appear to imply that the Internet 2
client is a dedicated desktop system running a multi-threaded,
multi-tasking operating system (NT or UNIX in today's terms) on
a high end processor (RISC or Pentium Pro in today's terms) with
a high bandwidth connection (e.g., at least 25 Mbps). This indeed
may be the dominant, but not exclusive, platform. Very soon, the
term "desktop computing" may become an anachronism with
the explosion of other types of communication devices. Therefore,
the Internet 2 applications environment must work within a mesh
of connectivity where an individual with multiple access devices
receives communicates across a complex mesh of networks. This
world of devices could include personal digital assistants, laptops,
and fixed workstations with overlapping functionality; PDAs and
portable phones blurring in a PCS world; "set-top" boxes
(e.g., WebTV) providing functionality that competes with the PC;
and networks of embedded systems supporting applications from
the simple to the complex. All of this will be in a networked
environment with connectivity options including the highest direct
Internet 2 end-to-end connection to spread-spectrum wireless services
and everything in between.
A primary goal of the Internet 2 Applications Working Group is
to facilitate and coordinate the creation of an applications architecture
and application development tools to inform and take advantage
of Internet 2 advanced network services. These tools are most
likely to arise in the process of developing specific applications
across a range of application areas, but their ultimate value
will be to seed the long-term distributed development of applications
to support higher education's mission of instruction, research,
and public service.
Some application areas and a few related tools are outlined in
this concluding section to portray the nature of the advanced
applications that should drive the engineering of Internet 2.
There is very little high quality instructional software available
to serve as the content basis for distributed instruction. Most
instructional software has been designed for stand-alone use,
especially that which incorporates sound, image or video. Much
of this is dependent on a single operating system. Internet 2
is an opportunity to work on an applications development architecture
for learningware and applications related to its delivery and
use in distributed instruction.
Component technologies -- building blocks -- can encourage a "1,000
flowers to bloom." The foundations for such building blocks
are now emerging from the information technology industry in the
form of object-oriented development tools and distributed object
architectures -- DSOM, Java, Active-X, OpenDoc, to name a few.
These generic tools and "standards" will not provide
all of the building blocks necessary to create a distributed environment
for instruction and research even though they likely will solve
many problems -- authentication, authorization, and security,
for example. The new component tools and models, however, can
be extended to include the required functionalities. Creating
networked content materials for learning, for example, will be
much easier if generic, cross-platform building blocks and protocols
are available to developers. For example, the developer of an
application designed to allow students to collect and analyze
data from a scientific instrument on the Internet should have
access to a networked data sampling tool which recognizes various
data protocols, an intelligent plotting window with a variety
of display and scaling features, and a tool for passing sampled
data to the plotting window. With such tools, the developer could
concentrate on developing a networked learning environment incorporating
interactive data collection and analysis. More generally, the
interoperable building blocks required by content developers would
include 2- and 3-dimensional graphing templates, mathematical
modeling templates, symbolic computation and manipulation engines,
a mathematical scripting language, molecular modeling templates,
intelligent periodic tables and atomic bonding tools, other science-specific
generic functionalities, templates for developing case studies
around video clips, tools for glossing texts, tools for synchronizing
temporal data (such as music) with related text and images (such
as musical scores), bilingual lexical databases and search tools
for developing second-language acquisition applications, and many
other generic functionalities. These are the kind of building
blocks which can be the foundation for content for the Instructional
Management System described below.
Any instructional process, whether in a K-12, collegiate or training
environment, typically incorporates the following actions:
In the traditional instructional environment, this process is
designed, managed, and implemented by teachers. In networked,
distributed instructional environment, this process should be
designed by teachers but managed by software, and often should
be shared between teachers, students, and other entities such
as publishers and information providers. This network-based, instructional
management system is called the IMS. The IMS consists of both
standards and services. The standards will permit distributed
instructional modules to interoperate with regard to such aspects
as tracking of student progress, automated incorporation of modules
into broader frameworks, collaborative interaction, and flow between
modules. Standards will also create a common mechanism for organization
and retrieval of network-based instructional objects by reflecting
the relationship of individual instructional modules to specific
learning objectives. While some of the technologies of the IMS
could be developed in today's Internet environment, the synchronous
communication components and technologies for linking and delivering
multimedia-rich learning materials will require network services
not yet available.
EDUCOM's National Learning Infrastructure Initiative will create
and publish the IMS standards. The intent is for these standards
to be made widely available, so that commercial developers can
create proprietary IMS systems based on the generic standards,
in a manner that parallels the development and adoption of the
URL, HTML, HTTP standards in the context of the Web. Developers
of instructional modules will be able to use the standards as
a means of ensuring that their software modules will comply with
the IMS, regardless of the specific IMS implementation that is
used to manage them. The standards will define data elements that
will be incorporated into all IMS-compliant objectives, diagnostic,
and instructional modules, and will cover areas including related
objectives, learning styles, bookmarks, state information about
collaborative tools, etc.
Instructional modules will provide status reporting at a frequency
specified by the instructor or in response to an event within
the system, such as a student terminating a module. Modules will
report a variety of information including test results, dwell
time, and bookmark information. Modules will operate in various
modes:
Modules will also be able to receive management commands: e.g.
resume at a bookmark, set behavior meta data (when to signal the
management system after some period of student inactivity, etc.),
receive remote control commands and collaborative interaction
commands, and call other modules or utilities that enable the
primary module to fulfill instructional and system goals.
The Instructional Management System (IMS) initiative was designed
to address fundamental impediments in the growth of Internet-based
distributed learning identified through national efforts undertaken
as part of EDUCOM's National Learning Infrastructure Initiative.
EDUCOM continues as the consortial focal point for IMS activities.
California State University (CSU), Miami-Dade Community College,
the University of Michigan, and the University of North Carolina
at Chapel Hill (UNC) have been responsible for the design and
implementation of the IMS and will continue their collaboration
with Cal State taking the project lead. Internet 2 member institutions
may wish to contribute to this effort under Cal State's lead.
The study and practice of music provides a good example. Interesting
examples of learningware for the appreciation of music have been
developed at several institutions. Migrating exemplars, such as
those developed at Indiana University Purdue University at Indianapolis,
to a web-based environment is constrained by current limitations
on the quality of streaming audio. Internet 2 services could remove
these constraints, and the IMS could help instructors locate such
materials and utilize them in a distributed instructional environment
enhanced by a variety of synchronous and asynchronous tools for
student-instructor communication.
In an Internet 2 environment, moreover, studio instruction in
music also would have new opportunities. World-class musicians
could be invited to offer their insight and expertise. For example,
a two-way video/audio connection might link a high school jazz
band with an artist-in-residence at a university. The high quality
of the communication link would allow demonstration and critical
review to occur. In addition, the students would literally be
able to "jam" with the University-based instructor.
This connection could be extended to musicians (whether students
or professional artists) at additional locations. The instruction
could be enriched by introducing recorded audio and video performances
drawn from a network-based server. The student interaction with
the instructor could be recorded for later review, either by the
instructor or for practice by the students.
Current research efforts have already demonstrated that the existing
commodity Internet can be an effective environment for developing
digital library systems. These efforts include the ARPA/NASA/NSF-sponsored
Digital Library Programs, as well as the wide range of operational
institutional library systems offering access to online catalogs,
abstracting and indexing databases, and primary content, such
as journals in electronic formats. While today's operational systems
suffer from reliability and performance problems as a result of
shortcomings in the existing Internet, they do not call for substantially
higher application-dedicated bandwith or bandwidth reservation.
They require only that the existing Internet function smoothly
and reliably within its current design parameters. Moreover, many
of the hardest problems -- intellectual property rights and rights
management, and viable economic models for scholarly publishing
in the 21st centuryóare far beyond the scope of any networking
infrastructure program.
But the new services and capabilities envisioned for Internet
2 offer important opportunities to move the Digital Libraries
program into new areas. Very high bandwidth and bandwidth reservation
will allow currently exotic materials such as continuous digital
video and audio to move from research use (such as in the Carnegie-Mellon
University Digital Library Project) to much broader use. Images,
audio and video can, at least from a delivery point of view, move
into the mainstream currently occupied almost exclusively by textual
materials. This will also facilitate more extensive research in
the difficult problems of organizing, indexing, and providing
intellectual access to these classes of materials.
Just as operational digital libraries today are dominated by textual
material, the interface to information retrieval systems remains
primarily textual. Even in the web environment, interfaces are
textual, though perhaps enhanced with modest graphical or tabular
materials. While language and, hence, text continue to be central
tools in retrieving information, there is a substantial body of
research on information visualization that has come out of organizations
such as Xerox PARC over the last decade. This research promises
substantial help to users in organizing, navigating, and comprehending
large complex information spaces. These techniques use complex,
high-resolution graphics and animation to provide visual representations
of large amounts of textual information in much the same way that
supercomputer based visualization has helped scientists over the
past decade to gain new insights into large numeric datasets and
simulation outputs. Internet 2 should provide sufficient performance
to the desktop to permit information visualization technologies
to be evaluated in broad-based information retrieval applications.
Other capabilities of Internet 2, such as the ability to provide
real-time help or expert consultation via audio or video conferencing
as part of a user interface, also offer opportunities to enrich
and extend the current state of the art in information access
and retrieval systems.
Finally, the availability of ubiquitous multicasting capabilities
in Internet 2, combined with high reliability and the ability
to manage quality of service on large numbers of low-speed connections
will have important, though currently hard to predict, implications
for information distribution and the management of distributed
databases. Current Web-based systems such as Pointcast hint at
what may be possible. In Internet 2, it should be possible to
stream information of all types -- database updates, publication
announcements, telemetry, sensor readings -- to communities of
interested receivers, rather than having those sites periodically
query centralized databases for the latest information. It is
easy to envision financial telemetry or news wires moving to such
a mode of distribution, for example; but these are only the most
obvious and simple cases of what may become a fundamentally new
model for information distribution. It will be important, early
in the development of Internet 2, to "seed" the exploration
of this model by ensuring the availability of a number of interesting
data "channels." Considerable work will also be needed
to translate research work in reliable multicast protocols into
common operation in Internet 2; it will be desirable to ensure
that these are part of the common protocol infrastructure much
as the Transmission Control Protocol (TCP) serves as the common
infrastructure for reliable point-to-point data interchange in
the current commodity Internet.
Another implication of the availability of quality-of-service
controls and multicast is that Internet 2 will be far more hospitable
than today's Internet to connecting very large numbers of sensors.
Indeed, over time sensors might well outnumber workstations. The
capacity to make large amounts of "public" shared sensor
telemetry available to the Internet 2 community represents an
exciting opportunity to explore new classes of applications.
Tele-immersion is the effective combination of:
This combination, we believe, offers a new paradigm for human
communications and collaboration.
Tele-immersion has the potential to significantly change educational,
scientific and manufacturing paradigms. A tele-immersion system
would allow individuals at different locations to share a single
virtual environment. For example, participants would interact
with a virtual group at a conference table approximating what
would be possible in a physical room. The individuals could share
and manipulate data, simulations and models of molecular, physical
or economic constructs, and jointly participate in the simulation,
design review or evaluation process. As an example, consider mechanical
engineering students or industrial engineers working together
to design a new bridge or robot arm via tele-immersion. Group
members would be able to interact with other group members while
sharing the virtual object being modeled.
Tele-immersion applications require advances in Internet infrastructure
because of their high bandwidth, low latency, and time-dependent
synchronous communications characteristics. Without high speed
networks embodying advanced protocols, such as RSVP and Multicast,
the potential of tele-immersion applications to further education,
advance science and reduce the design cycles for many manufacturing
applications will never be realized.
A well coordinated research and development effort is needed on
a number of fronts. Tele-immersion applications will require significant
extensions to current cave technology in the areas of tracking,
rendering, human interfaces that enhance the shared presence and
manipulation experience, as well as shared work tools for communications
and collaboration. Very important will be the integration of real
images into virtual environments to enable the simulation of realistic
shared presence.
A great deal of work will also be required in such areas as image
processing/construction, sensory simulation (e.g., touch),
and synchronization of inputs and human responses from participating
geographically distributed caves. These areas will require very
low latency networks, and drive other network parameters. In addition,
if not designed carefully at the outset to make the best computing,
storage and communications tradeoffs, tele-immersion applications
could be extraordinarily bandwidth intensive and, thus, of limited
utility.
During July and August of 1996, relevant work in this area was
surveyed, including visits to a number of sites. In consultation
with some leading researchers, a small workshop was planned and
convened to bring together leaders representing key tele-immersion
components and issues such as hardware, software, and networking
and social implications.
The tele-immersion workshop was held in October of 1996 in Chicago,
and the key technology issues associated with evolving current
virtual reality system, such as caves into two- or multi-way collaborative
environments, were identified. Participants endorsed the goal
of establishing a program to bring together the requisite researchers
and skills to create an environment in which individuals in three
geographically distributed caves can experience the shared presence
of each other while examining and manipulating the results of
a computer generated model or simulation. The tele-immersion applications
supporting this environment would first focus on education, science,
and manufacturing.
A detailed plan, in the form of a proposal, is now being developed
outlining the specifics of what has to be done to build such a
three location tele-immersion environment and to locate the skills
needed for the effort. A proposed budget and management structure
for the project is also being prepared. The detailed plan will
provide a road map for what needs to be done, a budget to do it
and a management structure to ensure successful implementation.
This plan should be completed by December.
A Virtual Laboratory is a heterogeneous, distributed problem solving
environment that enables a group of researchers located around
the world to work together on a common set of projects. As with
any other laboratory, the tools and techniques are specific to
the domain of the research, but the basic infrastructure requirements
are shared across disciplines. Although related to some of the
applications of tele-immersion, the virtual laboratory does not
assume a priori the need for a shared immersive environment.
The Grand Challenge Computational Cosmology Consortium is a group
of theoretical astronomers and computer scientists that are engaged
in collaborative research on the origin of the Universe and the
emergence of large scale structures. This group includes scientists
from Indiana University, NCSA, Princeton, MIT, UC-SC and the Pittsburgh
Supercomputer Center. Their work requires massive simulations
that involve multiple supercomputers working together, large data
bases of simulation results, extensive visualizations that demonstrate
the evolution of stars and galaxies, and a large repository of
shared software to make this all happen. While some experiments
are conducted by individuals, the largest require the close collaboration
of teams of people distributed over multiple time zones. Each
team member is an expert on a particular component of the heterogeneous
mix of simulation, data analysis and visualization. The team must
be able to share a common view of the simulation and interactively
steer the collective computation.
As another example, consider multi-disciplinary design and manufacturing.
In this case, a company involved with the production of a large
and complex product, such as an airplane, must be able to direct
simulation processes to interact with design data bases which
contain technical and manufacturing specifications. The design
and simulation can require the simultaneous access to hundreds
of subcomputations which are provided by subcontractors at different
locations. The result is a "multi-disciplinary optimization"
where the most cost effective and safest product can be manufactured
to customer specifications.
A third example is a weather forecasting system that incorporates
satellite data, large numbers of sensor inputs, and massive simulations
for short and medium range weather predictions. A variation on
this is predicting air quality through a virtual laboratory that
couples weather models with ocean circulation models and pollution
chemistry with ground and air based particulant sensors. In such
a laboratory, an environmental scientists could suggest, given
current conditions, when to shut down temporarily certain types
of manufacturing to avoid a potential crisis in air quality. Virtual
laboratories have been proposed in many other disciplines including
computational biology, radio astronomy, drug design, and materials
science.
The components of a virtual laboratory include:
Tightly coupled, multi-disciplinary computations place great stress
on network bandwidth. Low latency is critical and computer system
resource scheduling must be coupled with bandwidth reservation
services. Multicast protocols and technology are critical to the
collaborative nature of an experiment in a virtual laboratory
where people, resources, and computations are widely distributed.
Information streams in these experiments might combine voice,
video, real-time data streams from instruments, and large bursts
of data from simulations and visualization sources.
The I-Way experiments at Supercomputing-95 provided the first
national scale test of an infrastructure to support virtual laboratories.
The results of this activity proved that the concept is feasible
and that it is possible to achieve real scientific goals in such
an environment. However, the I-Way network was very fragile and
the experiments also exposed critical weaknesses in the basic
software infrastructure for building distributed applications.
As a result of the I-Way work, several new projects have begun
to address application level software infrastructure. These projects
include the ARPA Globus project, the DOE Legion, and the Gigabit
CORBA work. Also under development are a number of programming
tools that use this emerging infrastructure to help programmers
design and build the applications that will run on the Internet
2. These tools range from network resource management and operating
systems schedulers to distributed objects systems that allow current
client-server models to scale to the level needed for the computations
described above.
Through a series of planned collaborations between government
labs, NSF programs, and industry and university research projects,
the software infrastructure for building virtual laboratories
could evolve with the development of Internet 2 over the next
few years.
Higher education was pervasively present at the birth of the Internet and has contributed significantly to the infrastructure and application base at the heart of the Internet's present success. It would be ironic if higher education failed to be a proactive participant in the evolution of the next generation of network applications which can be glimpsed on the Internet horizon today and which promise to alter radically the prevailing methodologies of instruction, research, and public service. The Internet 2 Project is a clear signal that higher education intends to contribute to the advance of those network technologies and, especially, those network applications which will be the foundation for the knowledge society envisioned in the Clinton-Gore Next Generation Internet Initiative. This document is a first stake in that applications ground.